JP7518541B2 - Cattle evaluation device and cattle evaluation method - Google Patents

Description

The present invention relates to a cow evaluation device and a cow evaluation method.

A method using rumen size score as a method for evaluating the health of cows is known (e.g., Non-Patent Document 1). In the method described in Non-Patent Document 1, veterinarians and dairy experts determine the rumen size score using a judgment sheet. With this method, each veterinarian or expert can obtain a stable evaluation of the rumen size score, but there is variation in the rumen size score between veterinarians and experts. On the other hand, methods are known that evaluate the health of cows by image analysis using an RGB camera (e.g., Patent Document 1) and that evaluate the health of cows based on three-dimensional images of the cow (e.g., Patent Document 2).

JP 2015-173732 A International Publication No. 2017/187719

"Application of rumen size score in dairy herd management," by Haruhisa Futamura, Gyuho Magazine, 2006, No. 34, pp. 44-48

The method described in Patent Document 1 evaluates the health of cows using line drawings that connect feature points on an image of the cow. In the method described in Patent Document 1, a person inputs the line drawings on the image, so there is a risk that the health evaluation results will vary depending on the person who inputs them. In addition, the method described in Patent Document 2 calculates a score indicating the health of the cow using a three-dimensional image of the cow, so it requires imaging at a specific location where large-scale equipment is installed, making it difficult to apply to a variety of breeding sites.

The present invention has been made in consideration of the above problems, and aims to provide a cow evaluation device and a cow evaluation method that can accurately determine a cow's rumen size score using two-dimensional captured images.

The cattle evaluation device of the present invention includes an image acquisition unit that acquires a two-dimensional image of a standing cattle taken from approximately directly behind, an extraction unit that extracts the contour line of the cattle from the captured image, a dimension determination unit that determines a first dimension indicating the size of the cattle's body in the vertical direction from the captured image using a predetermined method, a line determination unit that determines at least one of two points on the contour line located approximately on the left and right sides of a first point that is located downward from a reference point near the left-right center and near the top end of the cattle by a second dimension shorter than the first dimension, using a predetermined method, and determines at least one of a first line passing through one of the two points and the reference point, and a second line passing through the other point and the reference point, an angle acquisition unit that acquires the angle between the reference line identified from the captured image and the first line or the second line, or the angle between the first line and the second line, and a score determination unit that determines the rumen size score of the cattle from the angle acquired by the angle acquisition unit.

According to the present invention, it is possible to accurately determine a cow's rumen size score using two-dimensional captured images.

FIG. 1 is a block diagram showing the functional configuration of a cattle evaluation device according to the first embodiment. FIG. 2 is a block diagram showing the functional configuration of the control unit in the first embodiment. FIG. 3 is a diagram showing an example of an image acquired by the image acquisition unit. FIG. 4 is a diagram for explaining the process executed by the extraction unit. FIG. 5 is a diagram for explaining the processes executed by the reference point specifying unit and the dimension specifying unit in the first embodiment. FIG. 6 is a diagram for explaining the processes executed by the line determination unit and the angle acquisition unit in the first embodiment. 7(a) and 7(b) are images of cows with different degrees of ruminal fullness. 8(a) and 8(b) are diagrams showing the relationship between the degree of fullness of the first stomach (rumen) and the angle α acquired by the angle acquisition unit. FIG. 9 is an image of a cow with a low rumen size score captured from almost directly behind. FIG. 10 is an extracted image in which the contour line of the cow captured in the captured image of FIG. 9 is extracted. FIG. 11 is a diagram showing the results of an experiment investigating the correlation between the lumen size score and the angle α acquired by the angle acquisition unit. FIG. 12 is a diagram showing the correlation between the coefficients used by the calculation unit to calculate the dimensions and the coefficients of determination, obtained from the experimental results of FIG. FIG. 13 is a block diagram showing the hardware configuration of the cow evaluation device according to the first embodiment. FIG. 14 is a flowchart showing an example of a method for evaluating cattle according to the first embodiment. FIG. 15 is a diagram for explaining the processes executed by the reference specifying unit and the dimension specifying unit in the second embodiment. FIG. 16 is a diagram for explaining the processes executed by the line determination unit and the angle acquisition unit in the second embodiment. FIG. 17 is a diagram showing the results of an experiment investigating the correlation between the lumen size score and the angle β acquired by the angle acquisition unit. FIG. 18 is a diagram for explaining the processes executed by the dimension specifying unit, the straight line determining unit, and the angle obtaining unit in the third embodiment. FIG. 19 is a diagram showing the results of an experiment investigating the correlation between the lumen size score and the angles γ1 and/or γ2 acquired by the angle acquisition unit. FIG. 20 is a diagram for explaining a process executed by an angle acquisition unit in a modified example of the third embodiment. FIG. 21 is a diagram for explaining the processing executed by the reference point specification unit and the dimension specification in the fourth embodiment. FIG. 22 is a diagram for explaining the processes executed by the reference point specifying unit, the dimension specifying unit, the straight line determining unit, and the angle obtaining unit in the fourth embodiment. FIG. 23 is a diagram showing the results of an experiment investigating the correlation between the lumen size score and the angle θ acquired by the angle acquisition unit. FIG. 24 is a diagram showing the correlation between the coefficient used by the calculation unit to calculate the dimensions and the coefficient of determination, obtained from the experimental results of FIG. FIG. 25 is a diagram for explaining the processes executed by the straight line determination unit and the angle acquisition unit in the fifth embodiment. FIG. 26 is a diagram for explaining the processing executed by the straight line determination unit and the angle acquisition unit in the modified example of the fifth embodiment. FIG. 27 is a diagram showing the relationship between the distance between the apexes of the left and right hip angles of a cow and the cow's withers height. FIG. 28 is a diagram showing the results of an experiment investigating the correlation between the lumen size score and the angle β acquired by the angle acquisition unit.

First Embodiment
A cow evaluation device 100 according to a first embodiment will be described below with reference to Figs. 1 to 14. Fig. 1 shows in a block diagram the functional configuration of the cow evaluation device 100 according to the first embodiment. Fig. 2 shows in a block diagram the functional configuration of a control unit 15 in the first embodiment. In this first embodiment, the evaluation device 100 is a portable information device carried and used by an operator. Examples of portable information devices include smartphones, tablet computers, PDAs (Personal Digital Assistants), and smart glasses. The evaluation device 100 has, for example, a telephone function, an email function, a communication function for connecting to the Internet, and a data processing function for executing a program.

As shown in FIG. 1, the evaluation device 100 has a display unit 10, an input unit 11, a communication unit 12, an imaging unit 13, a memory unit 14, and a control unit 15.

The display unit 10 is a display device that displays images, various information, and input images such as touch operation buttons.

The input unit 11 is a device such as a touch panel or a switch. The touch panel accepts information input in response to a worker's touch and transmits the accepted information to the control unit 15. The touch panel is, for example, incorporated in the display unit 10. Therefore, the touch panel accepts various information inputs in response to the worker touching the surface of the display unit 10. The switch is an input member that accepts input from the worker to the evaluation device 100 and transmits the accepted information to the control unit 15.

The communication unit 12 is a communication interface that performs short-range wireless communication (e.g., NFC (Near Field Communication)) with other devices, and wireless communication (e.g., communication using a mobile phone line or a wireless LAN (Local Area Network)) with other devices connected to the network.

The imaging unit 13 is a device that captures images such as still images and videos. The imaging unit 13 captures images in response to an operator's operation of the input unit 11. For example, when identifying a cow's rumen size score, the operator uses the imaging unit 13 to capture an image of the cow in a standing position from approximately directly behind. Capturing an image from approximately directly behind means capturing an image that captures the contours of the cow's left and right hip corners and left and right abdominal areas, as well as the cow's feet. Therefore, the operator captures the cow from a position several meters (e.g., 1 m to 2 m) behind the cow, capturing an image of the cow up to the cow's feet and capturing the contours of the left and right hip corners and left and right abdominal areas.

The memory unit 14 stores various information, such as captured images of the cow, extracted images of the cow's contour lines, and evaluation results that identify the cow's rumen size score.

The control unit 15 is a processing unit that controls each unit of the evaluation device 100. As shown in FIG. 2, the control unit 15 has an image acquisition unit 20, an extraction unit 21, a reference point identification unit 22, a dimension identification unit 23, a calculation unit 24, a straight line determination unit 25, an angle acquisition unit 26, and a score identification unit 27.

Figure 3 shows an example of an image acquired by the image acquisition unit 20. As shown in Figure 3, the image acquisition unit 20 acquires an image 30 of a cow in an upright position, captured from approximately behind. The acquired image 30 may be an image captured by the imaging unit 13, an image received by the communication unit 12 from another device, or an image selected by the operator using the input unit 11 from among images stored in the memory unit 14. The image acquisition unit 20 may display the acquired image 30 on the display unit 10. This allows the operator to visually recognize on the display unit 10 the cow whose rumen size score is to be specified.

Figure 4 is a diagram for explaining the processing executed by the extraction unit 21. As shown in Figure 4, the extraction unit 21 extracts a contour line 40 that traces the outline of the cow from the captured image 30 of the cow acquired by the image acquisition unit 20. A commonly known image processing technique can be used to extract the contour line 40. The extraction unit 21 may display an extracted image 31 that shows only the extracted contour line 40 of the cow on the display unit 10. This allows the operator to visually recognize on the display unit 10 the outline of the cow from which the rumen size score will be determined. The outer shape of the extracted image 31 is the same as the outer shape of the captured image 30, and the position of the cow relative to the image frame is the same in the extracted image 31 and the captured image 30.

There are ridges 41a and 41b on both the left and right sides of the upper part of the contour line 40. The ridges 41a and 41b correspond to the cow's hip angles. The tips 43a and 43b at the lower part of the contour line 40 correspond to the hooves of the cow's hind legs.

Figure 5 is a diagram for explaining the processing executed by the reference point identification unit 22 and the dimension identification unit 23 in the first embodiment. As shown in Figure 5, in this first embodiment, the reference point identification unit 22 identifies the approximate midpoint of a line segment 50 connecting the vertex 42a of the mountain portion 41a on the left side of the upper part of the cow's contour line 40 and the vertex 42b of the mountain portion 41b on the right side as the reference point 60. The reference point 60 is a point located near the center in the left-right direction of the cow and near the upper end. In addition, since the anus is located near the center between the left and right hip angles, the reference point 60 is also a point located near the anus. Note that the reference point identification unit 22 may identify the reference point from the vertex 42a of the mountain portion 41a and the vertex 42b of the mountain portion 41b by another method.

The apex 42a of the mountain portion 41a and the apex 42b of the mountain portion 41b can be identified by the following method. For example, an inclined straight line 51 inclined from the top to the bottom left in the extracted image 31 may be moved from the upper left end point 32 of the extracted image 31 toward the contour line 40, and the point where the inclined straight line 51 first touches the contour line 40 may be identified as the apex 42a of the mountain portion 41a located on the left side. Similarly, an inclined straight line 52 inclined from the top to the bottom right in the extracted image 31 may be moved from the upper right end point 33 of the extracted image 31 toward the contour line 40, and the point where the inclined straight line 52 first touches the contour line 40 may be identified as the apex 42b of the mountain portion 41b located on the right side. The inclined straight lines 51 and 52 may be, for example, straight lines inclined at 30° to 60°, 40° to 50°, or 45° from the horizontal line in the extracted image 31. Additionally, the inclined straight lines 51 and 52 may be inclined at an angle such that the lower sides of the peaks 41a and 41b are identified as apexes 42a and 42b.

The apex 42a of the mountain portion 41a and the apex 42b of the mountain portion 41b may be identified by other methods, not limited to the above-mentioned method. For example, a first point of contact may be identified by approaching a slanted straight line inclined from the top to the bottom left in the extracted image 31 from the upper left end point 32 toward the contour line 40, a second point of contact may be identified by approaching a slanted straight line inclined at a second predetermined angle from the upper left end point 32 toward the contour line 40, and a point located approximately in the center between the first point and the second point on the contour line 40 may be identified as the apex 42a of the mountain portion 41a. Similarly, a third point where an inclined line inclined from the top to the bottom right in the extracted image 31 at a third predetermined angle is first contacted by approaching the contour line 40 from the upper right end point 33, and a fourth point where an inclined line inclined at a fourth predetermined angle is first contacted by approaching the contour line 40 from the upper right end point 33, and a point on the contour line 40 located approximately in the middle between the third point and the fourth point is identified as the apex 42b of the mountain portion 41b. The first and third predetermined angles may be 15° to 30° from the horizontal line in the extracted image 31, and the second and fourth predetermined angles may be 60° to 75° from the horizontal line in the extracted image 31.

For example, the extracted image 31 may be converted into an X-Y coordinate system, and the vertices 42a, 42b of the two peaks 41a, 41b may be identified from the X and Y coordinates of each pixel of the extracted image 31. The vertex 42a of the peak 41a may be the point on the contour 40 that is the shortest distance from the upper left corner 32 of the extracted image 31, and the vertex 42b of the peak 41b may be the point on the contour 40 that is the shortest distance from the upper right corner 33 of the extracted image 31.

Alternatively, the apex of either the peak 41a or the peak 41b may be identified by any of the above methods, and the point where an approximately horizontal line in the extracted image 31 passing through the identified apex intersects with the contour line 40 of the other peak may be identified as the apex of the other peak.

As shown in FIG. 5, in this first embodiment, the dimension specification unit 23 first specifies, as a straight line 53, an approximately horizontal line in the extracted image 31 that passes through the lower end 43a of the two ends 43a, 43b of the contour line 40. In other words, the dimension specification unit 23 specifies, as a straight line 53, an approximately horizontal line in the extracted image 31 that passes through the lowermost end 43a of the contour line 40. Next, the dimension specification unit 23 specifies the distance L1 between the reference point 60 and the straight line 53 as a dimension that indicates the size of the cow's body in the vertical direction. In other words, the dimension specification unit 23 specifies the distance L1 between the reference point 60 and the intersection point 61 where the perpendicular line 59 in the extracted image 31 that passes through the reference point 60 intersects with the straight line 53 as a dimension that indicates the size of the cow's body in the vertical direction.

The calculation unit 24 calculates the dimension L2 by multiplying the dimension L1 identified by the dimension identification unit 23 by a predetermined coefficient. In this first embodiment, the predetermined coefficient is, for example, a value in the range of 0.2 to 0.4.

Figure 6 is a diagram for explaining the processing executed by the line determination unit 25 and the angle acquisition unit 26 in the first embodiment. As shown in Figure 6, in this first embodiment, the line determination unit 25 first identifies a point 62 that is located downward from the reference point 60 (e.g., directly below) by the dimension L2 calculated by the calculation unit 24, and identifies two intersections where an approximately horizontal line 54 in the extracted image 31 that passes through point 62 intersects with the contour line 40 as points 44 and 45. Next, the line determination unit 25 determines a line 55 that passes through point 44 and the reference point 60, and a line 56 that passes through point 45 and the reference point 60.

In the first embodiment, the angle acquisition unit 26 acquires the angle α between the straight lines 55 and 56. The angle α is, for example, an angle located above and below the reference point 60 among the multiple angles formed by the intersection of the straight lines 55 and 56.

The score determination unit 27 determines the rumen size score of the cow captured in the captured image 30 from the angle α acquired by the angle acquisition unit 26.

The rumen size score is an index of a cow's health, and evaluates the fullness of the first stomach (rumen). Cows with healthy first stomachs consume more dry matter, and therefore the first stomach becomes more swollen. As the first stomach is located in the left abdomen, the greater the swollen first stomach, the greater the left abdominal protrusion. The rumen size score evaluates the degree of this left abdominal protrusion from the relative shape of the entire abdomen when the cow is viewed from directly behind in an upright position.

7(a) and 7(b) are images 30 of cows with different degrees of fullness of the first stomach (rumen). The cow in the image 30 in FIG. 7(b) has a higher degree of fullness of the first stomach than the cow in the image 30 in FIG. 7(a). FIG. 8(a) and FIG. 8(b) are diagrams showing the relationship between the degree of fullness of the first stomach (rumen) and the angle α acquired by the angle acquisition unit 26. FIG. 8(a) and FIG. 8(b) show an extracted image 31 in which the outline 40 of the cow captured in the image 30 in FIG. 7(a) and FIG. 7(b) is extracted. As shown in FIG. 7(a) to FIG. 8(b), the higher the degree of fullness of the first stomach, the greater the degree of abdominal protrusion, and therefore the larger the angle α. Therefore, it is considered that the degree of fullness of the first stomach can be evaluated and the rumen size score can be determined by determining the angle α.

Here, the cow in the captured images 30 of Figures 7(a) and 7(b) has an abdomen that protrudes more outwardly to the left and right than other parts. For this reason, it is conceivable to use the point on the cow's contour that protrudes most outwardly to the left and right to determine the rumen size score. However, if the cow is thin, i.e., if the rumen size score is low, when the cow is photographed standing from almost directly behind, the buttocks may protrude more outward to the left and right than the abdomen. This will be explained using figures.

9 is a captured image 30 of a cow with a low rumen size score captured from almost directly behind. FIG. 10 is an extracted image 31 in which the contour line 40 of the cow captured in the captured image 30 of FIG. 9 is extracted. As shown in FIG. 9, in the case of a thin cow with a low rumen size score, the buttocks 91 may protrude outward to the left and right from the abdomen 90 where the stomach is located. In this case, as shown in FIG. 10, the leftmost point 93 and the rightmost point 94 of the contour line 40 correspond to the buttocks. Therefore, even if the rumen size score is determined using the angle formed by the line 95 passing through the leftmost point 93 and the reference point 60, and the line 96 passing through the rightmost point 94 and the reference point 60, an accurate rumen size score cannot be determined. From this, it can be seen that it is preferable to determine the position of the abdomen and obtain the angle from the line passing through the specified point in the abdomen and the reference point 60.

[Experiment 1]
An experiment was conducted to investigate the correlation between the rumen size score determined by an expert using the assessment sheet described in Non-Patent Document 1 and the magnitude of the angle α acquired by the angle acquisition unit 26. The experiment was conducted on dairy cows (Holstein breed) (the same applies to the following Experiments 2 to 5).

Figure 11 shows the results of an experiment investigating the correlation between the rumen size score and the angle α acquired by the angle acquisition unit 26. The horizontal axis (x-axis) of Figure 11 is the rumen size score determined by an expert using a judgment sheet, and the vertical axis (y-axis) is the angle α acquired by the angle acquisition unit 26. Each point in Figure 11, ○, △, □, ●, ▲, ■, and ×, represents each dairy cow in the experiment, and the straight lines represent approximate straight lines. In addition, ○, △, □, ●, ▲, ■, and × represent cases in which the coefficients used by the calculation unit 24 to calculate the dimension L2 are different from one another. The coefficients used to calculate the dimension L2 are ○: 0.15, △: 0.2, □: 0.25, ●: 0.3, ▲: 0.35, ■: 0.4, and ×: 0.45.

As shown in FIG. 11, when the coefficient used to calculate the dimension L2 was in the range of 0.15 to 0.45, the angle α increased as the lumen size score increased. The approximate line for the results of ○ was y = 5.64x + 96.99, and the coefficient of determination ^R2 was 0.389. The approximate line for the results of △ was y = 9.28x + 72.46, and the coefficient of determination ^R2 was 0.594. The approximate line for the results of □ was y = 12.90x + 50.55, and the coefficient of determination ^R2 was 0.724. The approximate line for the results of ● was y = 15.76x + 31.68, and the coefficient of determination ^R2 was 0.765. The approximate line for the results of ▲ was y = 13.96x + 27.13, and the coefficient of determination ^R2 was 0.717. The approximate line for the result of (2) was y = 10.11x + 27.50, and the coefficient of determination ^R2 was 0.618. The approximate line for the result of (x) was y = 5.52x + 31.51, and the coefficient of determination ^R2 was 0.341.

Figure 12 is a diagram showing the correlation between the coefficient used by the calculation unit 24 to calculate the dimension L2 and the coefficient of determination obtained from the experimental results of Figure 11. The horizontal axis of Figure 12 is the coefficient used to calculate the dimension L2, and the vertical axis is the coefficient of determination. As shown in Figure 12, the coefficient of determination was largest when the coefficient used to calculate the dimension L2 was 0.3. Furthermore, the coefficient of determination of about 0.6 or more was obtained by setting the coefficient to a value within the range of 0.2 to 0.4.

In this way, when the coefficient used to calculate dimension L2 is set within the range of 0.2 to 0.4, the correlation between the rumen size score and the angle α formed by the line 55 passing through point 44 and reference point 60 and the line 56 passing through point 45 and reference point 60 becomes stronger, so it can be said that points 44 and 45 are points located in the cow's abdomen. In other words, dimension L2 is calculated by multiplying dimension L1 specified by the dimension specifying unit 23 by a coefficient within the range of 0.2 to 0.4, and it can be said that the surface of the cow located approximately on both the left and right sides of the point that is located downward from reference point 60 by the calculated dimension L2 is the surface of the abdomen. This will be explained using Figure 9.

In FIG. 9, when the coefficient used by the calculation unit 24 to calculate the dimension L2 is 0.2, the horizontal line passing through a point that is distanced downward from the reference point 60 by the dimension L2 is indicated by dashed line 97, and when the coefficient is 0.4, the horizontal line passing through a point that is distanced downward from the reference point 60 by the dimension L2 is indicated by dashed line 98. As shown in FIG. 9, the dimension L2 is calculated using a coefficient in the range of 0.2 to 0.4, and the surface of the cow located approximately on both sides of the point that is distanced downward from the reference point 60 by the calculated dimension L2 falls within the range between dashed lines 97 and 98, and becomes the surface of the cow's abdomen. Therefore, by identifying points on the surface of the cow that are approximately on both sides of the point that is distanced downward from the reference point 60 by the dimension L2, it is possible to identify points located in the cow's abdomen, and as a result, it becomes possible to accurately identify the rumen size score from the angle α.

As such, since there is a strong correlation between the lumen size score and the angle α, the score determination unit 27 can determine the lumen size score from the angle α with high accuracy. The score determination unit 27 determines the lumen size score from the angle α, for example, by using score information that associates the angle with the lumen size score and that is stored in the memory unit 14.

Table 1 shows an example of score information stored in the storage unit 14. As shown in Table 1, in the score information, angles and lumen size scores are associated with each other. For example, when the angle is less than a°, the lumen size score is 1.00, and when the angle is equal to or greater than a° and less than b°, the lumen size score is 1.50. In Table 1, the lumen size scores are in increments of 0.25 between 2.00 and 4.00, and in increments of 0.50 otherwise, but other cases are also possible. For example, the increments may be 0.25 or 0.50 in the entire range.

The score information stored in the memory unit 14 may be other than the score information shown in Table 1, as long as it is information that associates the angle with the lumen size score. For example, a linear function (y = ax + b) corresponding to the approximation line shown in FIG. 11 may be stored in the memory unit 14 as score information. In this case, the lumen size score may be determined by performing fractional processing such as rounding up, rounding up, or rounding down on the value of x obtained by substituting the angle for y.

Figure 13 is a block diagram showing the hardware configuration of the cow evaluation device 100 according to the first embodiment. As shown in Figure 13, the cow evaluation device 100 has a storage device 70, a memory 71, a processor 72, a communication interface 73, an input device 74, a display device 75, and an imaging device 76. These components are interconnected by a bus 77.

The storage device 70 is a non-volatile semiconductor memory such as a flash memory, and stores the cow evaluation program 78.

Memory 71 is hardware that temporarily stores data, such as a dynamic random access memory (DRAM). The memory unit 14 in FIG. 1 is realized by the storage device 70 and memory 71.

The processor 72 is hardware such as a CPU (Central Processing Unit) that controls each part of the cow evaluation device 100. The processor 72 executes the cow evaluation program 78 in cooperation with the memory 71. In this way, the memory 71 and the processor 72 cooperate to execute the cow evaluation program 78, thereby realizing each function of the control unit 15 in FIG. 2.

The communication interface 73 is hardware that enables the cow evaluation device 100 to communicate with other devices by short-range wireless communication or wireless communication via a network. The communication interface 73 realizes the communication unit 12 in FIG. 1.

The input device 74 is an input device such as a touch panel or a switch that allows the operator to operate the cattle evaluation device 100. The input device 74 realizes the input unit 11 in FIG. 1.

The display device 75 is hardware for implementing the display unit 10 in FIG. 1, and is a display device such as an LCD display for displaying the captured image 30 of the cow, the extracted image 31, the rumen size score evaluation results, etc.

The imaging device 76 is hardware for implementing the imaging unit 13 in FIG. 1, and includes a lens, an imaging element, etc.

Next, an example of the cattle evaluation method according to the first embodiment will be described with reference to the flowchart in FIG. 14. The process in FIG. 14 is executed by the control unit 15 of the evaluation device 100. It is assumed that the memory unit 14 stores score information such as that shown in Table 1, for example.

In the process of FIG. 14, first, in step S10, the image acquisition unit 20 of the control unit 15 waits until the cow evaluation program 78 is started. In this case, the process moves to step S12 when the control unit 15 starts the cow evaluation program 78 in response to an operation by the operator (e.g., an input operation to the input unit 11 or voice input to a microphone (not shown)).

When the process proceeds to step S12, the image acquisition unit 20 acquires an image 30 of the cow for which the rumen size score is to be specified (see FIG. 3). The image 30 of the cow is an image of the cow in an upright position captured from approximately directly behind. The image acquisition unit 20 acquires, for example, an image of the cow captured by the operator using the imaging unit 13. The image acquisition unit 20 may acquire an image of the cow received from another device via the communication unit 12, or an image of the cow selected by the operator by operating the input unit 11 from among multiple images of the cow stored in the memory unit 14. The image acquisition unit 20 may start up a camera function so that the operator can capture an image of the cow using the imaging unit 13 after the cow evaluation program 78 is started. When the operator captures an image of the cow using the imaging unit 13, the image of the captured cow and a selection button asking whether or not to select this captured image may be displayed on the display unit 10.

In step S14, the extraction unit 21 of the control unit 15 extracts a contour line 40 that traces the outline of the cow from the captured image 30 of the cow acquired in step S12 (see FIG. 4). This generates an extracted image 31 showing the cow's contour line 40. When extracting the contour line 40, the extraction unit 21 may perform a process to remove background other than the cow from the captured image 30 in order to extract the contour line 40 with good accuracy.

Next, in step S15, the reference point identification unit 22 of the control unit 15 determines whether it is possible to identify the reference point 60 from the vertices 42a, 42b of the two peaks 41a, 41b that exist on both the left and right sides of the upper part of the contour line 40. If it is possible (Yes), the process proceeds to step S16, where the reference point 60 is identified from the vertices 42a, 42b of the two peaks 41a, 41b that exist on both the left and right sides of the upper part of the contour line 40 (see FIG. 5). For example, the reference point identification unit 22 moves the inclined straight lines 51, 52 that are inclined obliquely in the extracted image 31 closer to the contour line 40, identifies the points that first touch the contour line 40 as the vertices 42a, 42b of the peaks 41a, 41b, and identifies the approximate midpoint of the line segment 50 that connects the vertices 42a, 42b as the reference point 60.

In step S15, for example, if the reference point identification unit 22 determines that the captured image 30 of the cow is not suitable for identifying the rumen size score, such as when the line segment 50 is inclined at a predetermined angle or more relative to the horizontal line in the extracted image 31 or the frame of the extracted image 31, it determines that it is not possible to identify the reference point 60 (No). In this case, the process returns to step S12 and the process is repeated from the acquisition of the captured image 30. At this time, the reference point identification unit 22 may display a message on the display unit 10 or output it by voice, urging the operator to reacquire the captured image 30.

In step S18, the dimension determination unit 23 of the control unit 15 determines the distance L1 between the reference point 60 and a straight line 53, which is a substantially horizontal line in the extracted image 31 that passes through the tip 43a located at the bottom of the contour line 40, as a dimension indicating the size of the cow's body in the vertical direction (see Figure 5).

Next, in step S20, the calculation unit 24 of the control unit 15 calculates the dimension L2 by multiplying the dimension L1 identified by the dimension identification unit 23 by a predetermined coefficient. The predetermined coefficient is, for example, a value in the range of 0.2 to 0.4. Therefore, the dimension L2 is shorter than the dimension L1.

Next, in step S22, the straight line determination unit 25 of the control unit 15 identifies two points 44, 45 on the contour line 40 located on both sides of a point 62 that is located downward (e.g., directly below) by a dimension L2 from the reference point 60, and determines straight lines 55, 56 that pass through the points 44, 45 and the reference point 60 (see FIG. 6). For example, the straight line determination unit 25 identifies a point 62 that is located directly below the reference point 60 by a dimension L2, and identifies the two intersections where an approximately horizontal line 54 in the extracted image 31 that passes through the point 62 intersects with the contour line 40 as points 44, 45. The straight line determination unit 25 then determines a straight line 55 that passes through the point 44 and the reference point 60, and a straight line 56 that passes through the point 45 and the reference point 60.

Next, in step S24, the angle acquisition unit 26 of the control unit 15 acquires the angle α between the straight lines 55 and 56 determined in step S16 (see FIG. 6).

Next, in step S26, the score determination unit 27 of the control unit 15 determines the rumen size score of the cow captured in the captured image 30 from the angle α acquired in step S24 and the score information such as Table 1 stored in the memory unit 14. Next, in step S28, the score determination unit 27 displays the rumen size score determined in step S26 on the display unit 10 and stores it in the memory unit 14. For example, the score determination unit 27 may display the rumen size score on the display unit 10 together with the captured image 30 of the cow. In addition, the score determination unit 27 may store in the memory unit 14 as a single piece of information that associates the individual identification number of the cow for which the rumen size score has been determined, the captured image 30 of the cow, the image capture date, and the rumen size score of the cow.

As described above, according to the first embodiment, the image acquisition unit 20 acquires a two-dimensional captured image 30 of a cow in a standing position captured from approximately directly behind (see FIG. 3). The extraction unit 21 extracts the cow's contour line 40 from the captured image 30 (see FIG. 4). The dimension specification unit 23 specifies the dimension L1 (first dimension) indicating the size of the cow's body in the vertical direction from the captured image 30 using a predetermined method (see FIG. 5). The straight line determination unit 25 specifies points 44 and 45 on the contour line 40 located approximately on both the left and right sides of a point 62 located downward from a reference point 60 near the center in the horizontal direction of the cow and near the upper end by a dimension L2 (second dimension) shorter than the dimension L1, and determines straight lines 55 and 56 (first and second straight lines) passing through the points 44 and 45 and the reference point 60 (see FIG. 6). The angle acquisition unit 26 acquires the angle α between the straight lines 55 and 56 (see FIG. 6). The score determination unit 27 determines the rumen size score of the cow captured in the captured image 30 from the angle α acquired by the angle acquisition unit 26. In this way, in the first embodiment, the control unit 15 determines the straight lines 55 and 56, acquires the angle α between the straight lines 55 and 56, and determines the rumen size score from the angle α. This makes it possible to reduce the variation in the rumen size score compared to when a person determines the rumen size score as in Non-Patent Document 1. In addition, the points 44 and 45 on the contour line 40 determined by the straight line determination unit 25 are points located on both the left and right sides of the point 62 located below the reference point 60 by the dimension L2, and therefore are points located on the surface of the abdomen of the cow as described in FIG. 9. Therefore, by determining the rumen size score from the angle α between the straight lines 55 and 56, the rumen size score can be determined with high accuracy whether the cow is thin or fat. In addition, because the angle α does not change whether the cow is shown large or small in the captured image 30, which is a two-dimensional image, the rumen size score can be determined from the angle α, and therefore the rumen size score can be determined with high accuracy without performing processing such as normalization. Furthermore, because the rumen size score is determined using the two-dimensional captured image 30, which is captured from almost directly behind a cow in an upright position, the processing load is smaller than when the rumen size score is determined using a three-dimensional image of the cow, and the rumen size score can be determined even using a simple evaluation device 100 that is highly portable and versatile, such as a smartphone.

In the first embodiment, the straight line determination unit 25 identifies a point 62 (first point) that is separated downward from the reference point 60 by a dimension L2, and identifies two intersections where an approximately horizontal line 54 in the extracted image 31 that passes through the point 62 intersects with the contour line 40 as points 44 and 45 (see FIG. 6). When determining a straight line 55 (first straight line), the straight line determination unit 25 determines a straight line that passes through one of the two points 44 and 45, point 44, and the reference point 60 as the straight line 55 (see FIG. 6). When determining a straight line 56 (second straight line), the straight line determination unit 25 determines a straight line that passes through the other of the two points 44 and 45, point 45, and the reference point 60 as the straight line 56 (see FIG. 6). This determines straight lines 55, 56 passing through points 44, 45 located on the surface of the cow's abdomen and reference point 60, and the rumen size score of the cow can be determined with high accuracy by determining the angle α between straight lines 55 and 56. The approximately horizontal line 54 in the extracted image 31 does not have to be a completely horizontal line, and may be inclined to an extent that does not interfere with determining the rumen size score, or may be inclined within a range of ±5° from the horizontal line.

In the first embodiment, the reference point identification unit 22 identifies the approximate midpoint of the line segment 50 connecting the vertices 42a, 42b of the mountain portions 41a, 41b on both the left and right sides of the upper part of the contour line 40 as the reference point 60 (see FIG. 5). The dimension identification unit 23 identifies the distance L1 between the reference point 60 and a straight line 53, which is an approximately horizontal line in the extracted image 31 passing through the lower tip 43a of the two tips 43a, 43b present at the lower part of the contour line 40, as a dimension indicating the size of the body in the vertical direction of the cow (see FIG. 5). Then, the straight line determination unit 25 determines the straight lines 55 and 56 based on the items identified by the reference point identification unit 22 and the dimension identification unit 23, the angle acquisition unit 26 acquires the angle α between the straight lines 55 and 56, and the score identification unit 27 identifies the rumen size score from the angle α (see FIG. 6). As a result, the rumen size score is determined from the two-dimensional captured image 30 of the cow in a standing position, captured from almost directly behind, by the processing of the control unit 15 alone, without any human instruction, and the variation in the rumen size score can be suppressed to a small value. In addition, the reference point determination unit 22 determines the approximate midpoint of the line segment 50 connecting the apex 42a of the crest 41a and the apex 42b of the crest 41b as the reference point 60, so that the reference point 60 is a point located near the cow's anus. By using such a point located near the anus as the reference point 60, the rumen size score can be determined with high accuracy from the angle α formed by the straight lines 55 and 56. The approximate midpoint of the line segment 50 connecting the apexes 42a and 42b of the crests 41a and 41b is not limited to the exact midpoint of the line segment 50, but may be located near the anus, and may be located within a range of 10% of the length of the line segment 50 on either side of the exact midpoint of the line segment 50. The straight line 53 being an approximately horizontal line in the extracted image 31 does not necessarily have to be a completely horizontal line, but may be tilted to an extent that does not interfere with identifying the lumen size score, or may be tilted within a range of ±5° from the horizontal line.

In FIG. 5, line segment 50 is approximately horizontal in extracted image 31, but may not be horizontal depending on the positions of vertices 42a, 42b of peaks 41a, 41b. Line segment 50 may also deviate slightly in the vertical direction from FIG. 5 depending on how vertices 42a, 42b are determined. However, even if line segment 50 is slightly inclined, the position of reference point 60 hardly changes, and even if line segment 50 is slightly deviated vertically, the value of angle α hardly changes, so there is almost no effect on the accuracy of identifying the lumen size score.

In addition, in this first embodiment, the calculation unit 24 calculates the dimension L2 by multiplying the dimension L1 identified by the dimension identification unit 23 by a coefficient that is a value in the range of 0.2 to 0.4. That is, the dimension L2 is a value obtained by multiplying the dimension L1 by a coefficient that is 0.2 to 0.4. In the case where the reference point identification unit 22 identifies the approximate midpoint of the line segment 50 connecting the vertices 42a, 42b of the peaks 41a, 41b as the reference point 60, and the dimension identification unit 23 identifies the distance L1 between the reference point 60 and the straight line 53 passing through the tip 43a of the contour line 40 as the dimension indicating the body size in the vertical direction of the cow, the rumen size score can be accurately identified by setting the coefficient to a value of 0.2 to 0.4 as shown in FIG. 12. In order to improve the accuracy of identifying the rumen size score, the coefficient is preferably 0.25 to 0.35, more preferably 0.28 to 0.32, and even more preferably 0.3.

In the above first embodiment, the dimension specification unit 23 specifies the distance L1 between the reference point 60 and the straight line 53 passing through the lower tip 43a of the tips 43a and 43b of the contour line 40 as the reference dimension in the vertical direction of the cow, but this is not limited to this case. The dimension specification unit 23 may specify the distance between the reference point 60 and an approximately horizontal line in the extracted image 31 passing through the upper tip 43b of the tips 43a and 43b of the contour line 40, or a straight line passing through the tips 43a and 43b, as the dimension indicating the size of the body in the vertical direction of the cow. The dimension specification unit 23 may also specify the distance between the reference point 60 and an approximately horizontal line located midway between the approximately horizontal line in the extracted image 31 passing through the tip 43a and the approximately horizontal line in the extracted image 31 passing through the tip 43b, as the dimension indicating the size of the body in the vertical direction of the cow. Even in these cases, the dimensions indicating the size of the cow's body in the vertical direction do not change significantly, so the coefficient used by the calculation unit 24 should be 0.2 to 0.4, preferably 0.25 to 0.35, more preferably 0.28 to 0.32, and even more preferably 0.3.

In addition, in this first embodiment, the peaks 41a, 41b on both the left and right sides of the upper part of the contour line 40 correspond to the cow's hip angle. By identifying the reference point 60 from the vertices 42a, 42b (vertices of the parts corresponding to the cow's hip angle) of such peaks 41a, 41b, the reference point 60 will be located near the cow's anus, regardless of whether the cow is fat or thin. Therefore, the rumen size score can be identified with high accuracy.

Second Embodiment
The functional configuration and hardware configuration of the cattle evaluation device according to the second embodiment are the same as those of the first embodiment shown in Fig. 1 and Fig. 13. The functional configuration of the control unit 15 in the second embodiment is the same as that of the first embodiment shown in Fig. 2, but the processes executed by the reference point specifying unit 22, the dimension specifying unit 23, and the calculation unit 24 are different from those in the first embodiment. The processes executed by the reference point specifying unit 22, the dimension specifying unit 23, the calculation unit 24, the straight line determination unit 25, the angle acquisition unit 26, and the score specifying unit 27 in the second embodiment will be described below. The processes executed by the image acquisition unit 20 and the extraction unit 21 are the same as those in the first embodiment, and therefore will not be described.

15 is a diagram for explaining the processing executed by the reference point identification unit 22 and the dimension identification unit 23 in the second embodiment. As shown in FIG. 15, in this second embodiment, the reference point identification unit 22 identifies the point located at the top of the cow's contour line 40 as the reference point 60a. The reference point 60a is a point located near the center in the left-right direction of the cow and near the upper end. The reference point 60a is also a point located near the base of the cow's tail. In this second embodiment, the dimension identification unit 23 first identifies a straight line 53a that passes through the two tips 43a and 43b located at the bottom of the contour line 40. Then, the dimension identification unit 23 identifies the distance L3 between the reference point 60a and the straight line 53a as a dimension indicating the size of the cow's body in the vertical direction. That is, the dimension determination unit 23 determines the distance L3 between the reference point 60a and the intersection point 61a where the perpendicular line 59a in the extracted image 31 that passes through the reference point 60a intersects with the straight line 53a as the dimension that indicates the size of the cow's body in the vertical direction.

The calculation unit 24 calculates the dimension L4 by multiplying the dimension L3 identified by the dimension identification unit 23 by a predetermined coefficient. In this second embodiment, the predetermined coefficient is, for example, a value in the range of 0.3 to 0.4.

16 is a diagram for explaining the processing executed by the straight line determination unit 25 and the angle acquisition unit 26 in the second embodiment. As shown in FIG. 16, in the second embodiment, the straight line determination unit 25 first identifies a point 62a that is separated downward from the reference point 60a by the dimension L4 calculated by the calculation unit 24 (e.g., located directly below), and identifies two intersections at which the approximately horizontal line 54a in the extracted image 31 that passes through the point 62a intersects with the contour line 40 as points 44a and 45a. Next, the straight line determination unit 25 determines a straight line 55a that passes through the point 44a and the reference point 60a, and a straight line 56a that passes through the point 45a and the reference point 60a. In the second embodiment, the angle acquisition unit 26 acquires an angle β between the straight line 55a and the straight line 56a. The angle β is, for example, an angle located above and below the reference point 60a among a plurality of angles formed by the intersection of the straight lines 55a and 56a.

The score determination unit 27 determines the rumen size score of the cow captured in the captured image 30 from the angle β acquired by the angle acquisition unit 26.

The cattle evaluation method according to the second embodiment is similar to the flowchart of FIG. 14 of the first embodiment, except that the reference point 60a identified by the reference point identification unit 22, the dimension L3 identified by the dimension identification unit 23, and the dimension L4 calculated by the calculation unit 24 are different, and therefore will not be illustrated or described.

[Experiment 2]
An experiment was conducted to investigate the correlation between the rumen size score determined by an expert using the judgment sheet described in Non-Patent Document 1 and the magnitude of the angle β acquired by the angle acquisition unit 26. FIG. 17 shows the results of an experiment investigating the correlation between the rumen size score and the angle β acquired by the angle acquisition unit 26. The horizontal axis (x-axis) of FIG. 17 is the rumen size score determined by an expert using the judgment sheet, and the vertical axis (y-axis) is the angle β acquired by the angle acquisition unit 26. Each point of ○, △, and □ in FIG. 17 indicates each dairy cow in the experiment, and the straight line indicates an approximate straight line. In addition, ○, △, and □ indicate cases where the coefficients used by the calculation unit 24 to calculate the dimension L4 are different from each other. The coefficients used by the calculation unit 24 to calculate the dimension L4 are 0.3 for ○, 0.35 for △, and 0.4 for □.

As shown in Fig. 17, when the coefficient used by the calculation unit 24 to calculate the dimension L4 is in the range of 0.3 to 0.4, the larger the lumen size score, the larger the angle β. The approximate line for the results of ○ is y = 12.79x + 36.54, and the coefficient of determination ^R2 is 0.906. The approximate line for the results of △ is y = 13.24x + 26.47, and the coefficient of determination ^R2 is 0.881. The approximate line for the results of □ is y = 13.03x + 19.15, and the coefficient of determination ^R2 is 0.8571. In this way, by setting the coefficient used for calculating the dimension L4 within the range of 0.3 to 0.4, the correlation between the angle β formed by the line 55a passing through the point 44a and the reference point 60a and the line 56a passing through the point 45a and the reference point 60a and the rumen size score becomes stronger, so it can be said that the points 44a and 45a are points located in the abdomen of the cow. In other words, the dimension L3 specified by the dimension specifying unit 23 is multiplied by a coefficient within the range of 0.3 to 0.4 to calculate the dimension L4, and it can be said that the surface of the cow located approximately on both sides of the point located downward from the reference point 60a by the calculated dimension L4 is the surface of the abdomen. Therefore, by specifying the points on the surface of the cow located approximately on both sides of the point located downward from the reference point 60a by the dimension L4, it is possible to specify the points located in the abdomen of the cow, and as a result, it is possible to accurately specify the rumen size score from the angle β.

According to this second embodiment, the reference point identification unit 22 identifies the uppermost point of the contour line 40 as the reference point 60a (see FIG. 15). The dimension identification unit 23 identifies the distance L3 between the reference point 60a and the straight line 53a passing through the two tips 43a and 43b located at the bottom of the contour line 40 as the dimension indicating the size of the cow's body in the vertical direction (see FIG. 15). Even in such a case, the rumen size score can be accurately identified as described in FIG. 17.

In addition, in this second embodiment, the calculation unit 24 calculates the dimension L4 by multiplying the dimension L3 identified by the dimension identification unit 23 by a coefficient that is a value in the range of 0.3 to 0.4. That is, the dimension L4 is a value obtained by multiplying the dimension L3 by a coefficient that is 0.3 to 0.4. In a case where the reference point identification unit 22 identifies the uppermost point of the contour line 40 as the reference point 60a, and the dimension identification unit 23 identifies the distance L3 between the reference point 60a and the straight line 53a passing through the tip 43a and the tip 43b of the contour line 40 as a dimension indicating the body size in the vertical direction of the cow, the rumen size score can be accurately identified by setting the coefficient to a value of 0.3 to 0.4 as shown in FIG. 17. In order to improve the accuracy of identifying the rumen size score, the coefficient is preferably 0.3 to 0.38, more preferably 0.3 to 0.35, and even more preferably 0.3 to 0.32.

In the second embodiment, the dimension specification unit 23 is not limited to specifying the distance L3 between the reference point 60a and the straight line 53a passing through the tip 43a and the tip 43b of the contour line 40 as the dimension indicating the size of the body in the vertical direction of the cow, but may specify the distance between the reference point 60a and an approximately horizontal line in the extracted image 31 passing through one of the tips 43a and 43b of the contour line 40 as the dimension indicating the size of the body in the vertical direction of the cow. Even in this case, the dimension indicating the size of the body in the vertical direction of the cow does not change significantly, so the coefficient used by the calculation unit 24 may be 0.3 to 0.4, preferably 0.3 to 0.38, more preferably 0.3 to 0.35, and even more preferably 0.3 to 0.32. The coefficient used by the calculation unit 24 may be 0.2 to 0.4.

Third Embodiment
The functional configuration and hardware configuration of the cattle evaluation device according to the third embodiment are the same as those of the first embodiment shown in Fig. 1 and Fig. 13. The functional configuration of the control unit 15 in the third embodiment is the same as that of the first embodiment shown in Fig. 2, but the processes executed by the dimension specifying unit 23 and the angle acquiring unit 26 differ from those in the first embodiment. The processes executed by the dimension specifying unit 23, the calculation unit 24, the straight line determination unit 25, the angle acquiring unit 26, and the score specifying unit 27 in the third embodiment will be described below. The processes executed by the image acquiring unit 20, the extraction unit 21, and the reference point identifying unit 22 are the same as those in the first embodiment, and therefore will not be described.

Figure 18 is a diagram for explaining the processing executed by the dimension specification unit 23, the straight line determination unit 25, and the angle acquisition unit 26 in the third embodiment. Figure 18 shows the state after the reference point specification unit 22 specifies the reference point 60 using the same method as in the first embodiment. As shown in Figure 18, the dimension specification unit 23 specifies a straight line 53a that passes through the two tips 43a and 43b located at the bottom of the contour line 40. Then, the dimension specification unit 23 specifies the distance L5 between the reference point 60 and the straight line 53a as the dimension indicating the size of the cow's body in the vertical direction.

The calculation unit 24 calculates the dimension L6 by multiplying the dimension L5 identified by the dimension identification unit 23 by a predetermined coefficient. The predetermined coefficient is a value in the range of 0.2 to 0.4, as in the first embodiment.

The straight line determination unit 25 first identifies point 62b that is spaced downward from the reference point 60 by the dimension L6 calculated by the calculation unit 24 (e.g., located directly below), and identifies either point 44b or 45b that is the intersection point where a substantially horizontal line 54b in the extracted image 31 that passes through point 62b intersects with the contour line 40. Next, the straight line determination unit 25 determines either straight line 55b that passes through point 44b and the reference point 60, or straight line 56b that passes through point 45b and the reference point 60. Note that, for ease of understanding, in FIG. 18, both points 44b and 45b and both straight lines 55b and 56b are illustrated.

The angle acquisition unit 26 acquires either the angle γ1 between the reference line 57, which is a substantially perpendicular line in the extracted image 31 that passes through the reference point 60, and the straight line 55b, or the angle γ2 between the reference line 57 and the straight line 56b. The angle γ1 is, for example, an angle on the lower left or upper right side of the reference point 60 among the multiple angles formed by the intersection of the reference line 57 and the straight line 55b. The angle γ2 is, for example, an angle on the lower right or upper left side of the reference point 60 among the multiple angles formed by the intersection of the reference line 57 and the straight line 56b. Note that, for ease of understanding, both angles γ1 and γ2 are illustrated in FIG. 18.

The score determination unit 27 determines the rumen size score of the cow captured in the captured image 30 from the angle γ1 or γ2 acquired by the angle acquisition unit 26.

The cattle evaluation method according to the third embodiment is similar to the flowchart of FIG. 14 of the first embodiment, except that the dimension L5 determined by the dimension determination unit 23, the dimension L6 calculated by the calculation unit 24, the straight lines 55b and 56b determined by the straight line determination unit 25, and the angles γ1 and γ2 acquired by the angle acquisition unit 26 are different, and therefore will not be illustrated or described.

[Experiment 3]
An experiment was conducted to investigate the correlation between the rumen size score specified by an expert using the judgment sheet described in Non-Patent Document 1 and the angles γ1 and γ2 acquired by the angle acquisition unit 26. FIG. 19 shows the results of an experiment investigating the correlation between the rumen size score and the angles γ1 and/or γ2 acquired by the angle acquisition unit 26. The horizontal axis of FIG. 19 is the rumen size score specified by an expert using the judgment sheet, the right vertical axis is the angle γ1 or γ2 acquired by the angle acquisition unit 26, and the left vertical axis is the sum of the angles γ1 and γ2. Each point of ○, △, and × in FIG. 19 indicates each dairy cow in the experiment, and ○ indicates the result when the angle γ1 is acquired, △ indicates the result when the angle γ2 is acquired, and × indicates the result when the angles γ1 and γ2 are added together as a reference. In addition, the approximate straight line of the result of ○ is shown by a thick solid line, the approximate straight line of the result of △ is shown by a thin solid line, and the approximate straight line of the result of × is shown by a dashed line. In addition, in all of the cases of ◯, △, and ×, the coefficient used by the calculation unit 24 in calculating the dimension L6 was 0.3.

As shown in FIG. 19, the larger the lumen size score, the larger the angle γ1, the angle γ2, and the sum of the angles γ1 and γ2 all were. The coefficient of determination ^R2 of the approximation line of the result of ○ where the angle γ1 was obtained was 0.767, and the coefficient of determination R2 of the approximation line of the result of △ where the angle ^γ2 was obtained was 0.680. The coefficient of determination ^R2 of the approximation line of the result of × where the sum of the angles γ1 and γ2 was obtained for reference was 0.826. From this, it can be seen that the lumen size score can be determined with high accuracy even when the lumen size score is determined from the angle γ1 between the reference line 57 and the straight line 55b or the angle γ2 between the reference line 57 and the straight line 56b.

According to the third embodiment, the angle acquisition unit 26 acquires the angle γ1 between the reference line 57 and the straight line 55b, or the angle γ2 between the reference line 57 and the straight line 56b (see FIG. 18). The score determination unit 27 determines the rumen size score from the angle γ1 or γ2 acquired by the angle acquisition unit 26. Even in this case, as described in FIG. 19, the rumen size score can be determined with high accuracy. The angle acquisition unit 26 may acquire either the angle γ1 or the angle γ2, but since the first stomach (rumen) is located in the left abdominal region, it is preferable to acquire the angle γ1 from the viewpoint of determining the rumen size score with high accuracy. This can also be seen from the fact that, in FIG. 19, the coefficient of determination R ² of the approximation line of the result of △ in which the angle γ2 was acquired was 0.680, whereas the coefficient of determination R ² of the approximation line of the result of ◯ in which the angle γ1 was acquired was 0.767.

20 is a diagram for explaining the process executed by the angle acquisition unit 26 in the modified example of the third embodiment. As shown in FIG. 20, in the modified example of the third embodiment, the angle acquisition unit 26 acquires either the angle γ3 between the reference line 57a, which is a substantially horizontal line in the extracted image 31 passing through the reference point 60, and the straight line 55b, or the angle γ4 between the reference line 57a and the straight line 56b. The angle γ3 is, for example, an angle on the lower left or upper right side with respect to the reference point 60 among the multiple angles formed by the intersection of the reference line 57a and the straight line 55b. The angle γ4 is, for example, an angle on the lower right or upper left side with respect to the reference point 60 among the multiple angles formed by the intersection of the reference line 57a and the straight line 56b. Note that, in FIG. 20, both angles γ3 and γ4 are illustrated for ease of understanding.

In the third embodiment, the reference line 57 is an approximately perpendicular line in the extracted image 31. However, as in the modified example of the third embodiment, the reference line 57a may be an approximately horizontal line in the extracted image 31. By using the approximately perpendicular or horizontal lines in the extracted image 31 as the reference lines 57 and 57a, it is easy to determine the reference lines 57 and 57a, and it is also easy to obtain the angles γ1 and γ3 between the reference lines 57 and 57a and the straight line 55b and the angles γ2 and γ4 between the reference lines 57 and 57a and the straight line 56b. In the modified example of the third embodiment, the angles γ3 and γ4 obtained by the angle acquisition unit 26 become smaller as the degree of fullness of the first stomach (rumen) increases. The reference line 57 is not limited to being a completely perpendicular line in the extracted image 31, and may be inclined to a degree that does not interfere with the identification of the rumen size score, or may be inclined at about ±5° from the perpendicular line. Similarly, the reference line 57a being an approximately horizontal line in the extracted image 31 does not necessarily have to be a completely horizontal line, but may be tilted to an extent that does not interfere with the identification of the lumen size score, or may be tilted at about ±5° from the horizontal line. In addition, the reference line may be a straight line passing through the apex 42a of the peak 41a and the apex 42b of the peak 41b, or a straight line approximately perpendicular to the straight line passing through the apex 42a of the peak 41a and the apex 42b of the peak 41b.

In the third embodiment, the reference point identification unit 22 identifies the approximate midpoint of the line segment 50 connecting the vertex 42a of the crest 41a and the vertex 42b of the crest 41b as the reference point 60. However, as in the second embodiment, the reference point identification unit 22 may identify the uppermost point of the contour line 40 as the reference point 60a. Also, as in the first embodiment, the dimension identification unit 23 may identify the distance between the reference point 60 or 60a and the approximately horizontal line in the extracted image 31 that passes through one of the tips 43a, 43b located at the bottom of the contour line 40 as the dimension indicating the size of the cow's body in the vertical direction.

Fourth embodiment
In the first to third embodiments, the control unit 15 automatically identifies the reference points 60, 60a, etc. from the cow's contour line 40, but in the fourth embodiment, a case where the reference points are identified based on the operator's designation will be described. The functional configuration and hardware configuration of the cow evaluation device according to the fourth embodiment are the same as those in FIG. 1 and FIG. 13 of the first embodiment. The functional configuration of the control unit 15 in the fourth embodiment is the same as that in FIG. 2 of the first embodiment, but the processes executed by the reference point identification unit 22, the dimension identification unit 23, and the calculation unit 24 are different from those in the first embodiment. In the following, the processes executed by the reference point identification unit 22, the dimension identification unit 23, the calculation unit 24, the straight line determination unit 25, the angle acquisition unit 26, and the score identification unit 27 in the fourth embodiment will be described. The processes executed by the image acquisition unit 20 and the extraction unit 21 are the same as those in the first embodiment, so their explanations will be omitted.

Figure 21 is a diagram for explaining the processing executed by the reference point identification unit 22 and the dimension identification unit 23 in the fourth embodiment. Figure 22 is a diagram for explaining the processing executed by the reference point identification unit 22, the dimension identification unit 23, the straight line determination unit 25, and the angle acquisition unit 26 in the fourth embodiment. Figure 21 shows a captured image 30 of a cow acquired by the image acquisition unit 20, and Figure 22 shows an extracted image 31 in which the extraction unit 21 has extracted the cow's contour line 40. The captured image 30 in Figure 21 and the extracted image 31 in Figure 22 may be displayed side by side on the display unit 10, for example.

As shown in FIG. 21, the operator operates the input unit 11 to specify a location 80 near the left-right center and near the top end of the cow in the captured image 30 displayed on the display unit 10. The location 80 near the left-right center and near the top end of the cow is, for example, a location near the cow's anus. As shown in FIG. 22, the reference point identification unit 22 identifies a reference point 60b that corresponds to the location 80 near the left-right center and near the top end of the cow based on the operator's operation of the input unit 11.

When the worker operates the input unit 11 to specify the vertices of the left and right hip angles of the cow in the captured image 30 displayed on the display unit 10, the reference point identification unit 22 may identify the approximate midpoint of the line segment connecting the vertices of the left and right hip angles specified by the worker as the reference point 60b.

As shown in FIG. 21, the operator specifies a location 80 near the center and near the top of the cow in the captured image 30 displayed on the display unit 10, and then specifies the locations 81a and 81b of the cow's heel. As shown in FIG. 22, the dimension specification unit 23 specifies points 63a and 63b corresponding to the locations 81a and 81b of the cow's heel specified by the operator, and specifies a straight line 53b passing through the points 63a and 63b. Next, the dimension specification unit 23 specifies the distance L7 between the reference point 60b and the straight line 53b as a dimension indicating the size of the cow's body in the vertical direction. In other words, the dimension specification unit 23 specifies the distance L7 between the reference point 60b and the intersection point 61b where the perpendicular line 59b in the extracted image 31 passing through the reference point 60b intersects with the straight line 53b as a dimension indicating the size of the cow's body in the vertical direction.

The calculation unit 24 calculates the dimension L8 by multiplying the dimension L7 identified by the dimension identification unit 23 by a predetermined coefficient. In the fourth embodiment, the predetermined coefficient is, for example, a value in the range of 0.25 to 0.55.

As shown in FIG. 22, the straight line determination unit 25 first identifies point 62c that is located downward from reference point 60b (e.g., directly below) by the dimension L8 calculated by the calculation unit 24, and identifies points 44c and 45c that are intersections where a substantially horizontal line 54c in the extracted image 31 that passes through point 62c intersects with the contour line 40. Next, the straight line determination unit 25 determines a straight line 55c that passes through point 44c and reference point 60b, and a straight line 56c that passes through point 45c and reference point 60b. The angle acquisition unit 26 acquires the angle θ that is formed between straight lines 55c and 56c.

The score determination unit 27 determines the rumen size score of the cow captured in the captured image 30 from the angle θ acquired by the angle acquisition unit 26.

The cattle evaluation method according to the fourth embodiment is similar to the flowchart of FIG. 14 of the first embodiment, except that the reference point 60b identified by the reference point identification unit 22, the dimension L7 identified by the dimension identification unit 23, and the dimension L8 calculated by the calculation unit 24 are different, and therefore will not be illustrated or described.

[Experiment 4]
An experiment was conducted to investigate the correlation between the rumen size score determined by an expert using the judgment sheet described in Non-Patent Document 1 and the magnitude of the angle θ acquired by the angle acquisition unit 26. FIG. 23 shows the results of an experiment investigating the correlation between the rumen size score and the angle θ acquired by the angle acquisition unit 26. The horizontal axis (x-axis) of FIG. 23 is the rumen size score determined by an expert using the judgment sheet, and the vertical axis (y-axis) is the angle θ acquired by the angle acquisition unit 26. Each point of ○, △, □, ●, ▲, ■, ×, and ◇ in FIG. 23 indicates each dairy cow in the experiment, and the straight line indicates an approximation line. In addition, ○, △, □, ●, ▲, ■, ×, and ◇ indicate cases where the coefficients used by the calculation unit 24 to calculate the dimension L8 are different from each other. The coefficients used in calculating the dimension L8 are 0.25 for ○, 0.3 for △, 0.35 for □, 0.4 for ●, 0.416 for ▲, 0.45 for ■, 0.5 for ×, and 0.55 for ◇.

As shown in Fig. 23, when the coefficient used to calculate the dimension L8 was in the range of 0.25 to 0.55, the angle θ increased as the lumen size score increased. The approximate line for the results of ○ was y = 7.45x + 80.94, and the coefficient of determination ^R2 was 0.620. The approximate line for the results of △ was y = 11.11x + 61.57, and the coefficient of determination ^R2 was 0.827. The approximate line for the results of □ was y = 13.01x + 47.96, and the coefficient of determination ^R2 was 0.867. The approximate line for the results of ● was y = 13.73x + 37.98, and the coefficient of determination ^R2 was 0.848. The approximate line for the results of ▲ was y = 13.29x + 37.09, and the coefficient of determination ^R2 was 0.860. The approximate line for the results of (2) was y = 13.52x + 30.98, and the coefficient of determination ^R2 was 0.836. The approximate line for the results of × was y = 12.17x + 27.03, and the coefficient of determination ^R2 was 0.802. The approximate line for the results of ◇ was y = 8.57x + 29.57, and the coefficient of determination ^R2 was 0.700.

Figure 24 is a diagram showing the correlation between the coefficient used by the calculation unit 24 to calculate the dimension L8 and the coefficient of determination obtained from the experimental results of Figure 23. The horizontal axis of Figure 24 is the coefficient used to calculate the dimension L8, and the vertical axis is the coefficient of determination. As shown in Figure 24, the coefficient of determination was largest when the coefficient used to calculate the dimension L8 was 0.4. Furthermore, the results showed that a coefficient of determination of approximately 0.65 or more could be obtained by setting the coefficient to a value within the range of 0.25 to 0.55.

23 and 24, if the distance L7 between the reference point 60b and the straight line 53b passing through the points 63a and 63b corresponding to the cow's heels is identified as the dimension indicating the vertical body size of the cow, the identified dimension L7 is multiplied by a coefficient in the range of 0.25 to 0.55 to calculate the dimension L8, and it can be said that the surface of the cow located approximately on both sides of the point 62c located downward from the reference point 60b by the calculated dimension L8 is the surface of the abdomen. Therefore, by identifying points on the surface of the cow located approximately on both sides of the point located downward from the reference point 60b by the dimension L8, it is possible to identify points located in the cow's abdomen, and as a result, it is possible to accurately identify the rumen size score from the angle θ.

According to the fourth embodiment, the dimension specification unit 23 specifies the distance L7 between the reference point 60b and the straight line 53b passing through both heels of the cow's two hind legs as the dimension indicating the cow's body size in the vertical direction. Even in such a case, the rumen size score can be specified with high accuracy, as described in Figures 23 and 24.

In addition, in this fourth embodiment, the calculation unit 24 calculates the dimension L8 by multiplying the dimension L7 identified by the dimension identification unit 23 by a coefficient that is a value in the range of 0.25 to 0.55. That is, the dimension L8 is a value obtained by multiplying the dimension L7 by a coefficient that is 0.25 to 0.55. When the dimension identification unit 23 identifies the distance L7 between the reference point 60b near the cow's anus and the straight line 53b that passes through both heels of the cow's two hind legs as a dimension that indicates the body size in the vertical direction of the cow, the rumen size score can be accurately identified by setting the coefficient to a value of 0.25 to 0.55 as shown in FIG. 24. In order to improve the accuracy of identifying the rumen size score, the coefficient is preferably 0.28 to 0.52, more preferably 0.3 to 0.5, and even more preferably 0.35 to 0.45.

In the fourth embodiment, the dimension specification unit 23 is not limited to specifying the distance L7 between the reference point 60b and the straight line 53b passing through both heels of the cow as the dimension indicating the size of the body in the vertical direction of the cow, but may specify the distance between the reference point 60b and an approximately horizontal line in the captured image 30 passing through the heel of one of the two hind legs of the cow as the dimension indicating the size of the body in the vertical direction of the cow. Even in this case, the dimension indicating the size of the body in the vertical direction of the cow does not change significantly, so the coefficient used by the calculation unit 24 may be 0.25 to 0.55, preferably 0.28 to 0.52, more preferably 0.3 to 0.5, and even more preferably 0.35 to 0.45.

In the fourth embodiment, the operator specifies the heel of the cow, and the dimension specifying unit 23 specifies the distance between the reference point 60b and the straight line 53b passing through both heels of the cow as the dimension indicating the size of the body in the vertical direction of the cow. However, this is not limited to this case. The operator may specify the hoof of the cow's hind leg, and the dimension specifying unit 23 may specify the distance between the reference point 60b and an approximately horizontal line in the captured image 30 passing through one of the hooves of the cow's two hind legs, or a straight line passing through both hooves of the two hind legs, as the dimension indicating the size of the body in the vertical direction of the cow. The approximately horizontal line is not limited to a completely horizontal line, and may be inclined to an extent that does not interfere with the determination of the rumen size score, or may be inclined within a range of ±5° from the horizontal line.

In the fourth embodiment, the location 80 near the left-right center and near the top end of the cow specified by the operator is near the cow's anus, but it may be another location. For example, it may be near the base of the cow's tail. The angle acquisition unit 26 may also acquire the angle between the reference lines 57, 57a and the straight line 55c, or the angle between the reference lines 57, 57a and the straight line 56c, as in the third embodiment and its modified examples.

Fifth embodiment
In the first to fourth embodiments, a case has been described in which the straight line determination unit 25 determines a straight line that passes through the reference point and two points that are intersections at which an approximately horizontal line that passes through a point that is separated downward from the reference point by the dimension calculated by the calculation unit 24 intersects with the contour line 40. In the present fifth embodiment, a case will be described in which the straight line determination unit 25 specifies a range of interest that extends in the left-right direction in the extracted image 31 using a point that is separated downward from the reference point by the dimension calculated by the calculation unit 24, and determines straight lines that pass through the reference point and the leftmost and rightmost points on the contour line 40 that are located in the specified range of interest.

The functional configuration and hardware configuration of the cattle evaluation device according to the fifth embodiment are the same as those of the first embodiment, shown in Figs. 1 and 13. The functional configuration of the control unit 15 in the fifth embodiment is the same as that of the first embodiment, shown in Fig. 2, but the processes performed by the calculation unit 24 and the straight line determination unit 25 differ from those in the first embodiment. The processes performed by the calculation unit 24, straight line determination unit 25, angle acquisition unit 26, and score determination unit 27 in the fifth embodiment are described below. The processes performed by the image acquisition unit 20, extraction unit 21, reference point determination unit 22, and dimension determination unit 23 are the same as those in the first embodiment, and therefore will not be described.

Figure 25 is a diagram for explaining the processing executed by the straight line determination unit 25 and the angle acquisition unit 26 in the fifth embodiment. First, the calculation unit 24 calculates two dimensions L21 and L22 by multiplying the dimension L1 identified by the dimension identification unit 23 by different coefficients in the range of 0.2 to 0.4. For example, the coefficient used to calculate the dimension L21 is preferably a value of 0.2 to less than 0.3, more preferably a value of 0.2 to 0.25, and even more preferably a value of 0.2 to 0.22. The coefficient used to calculate the dimension L22 is preferably a value of more than 0.3 to 0.4, more preferably a value of 0.35 to 0.4, and even more preferably a value of 0.38 to 0.4.

25, the straight line determination unit 25 specifies a point 62d that is separated downward from the reference point 60 by the dimension L21 calculated by the calculation unit 24 (e.g., directly below), and a point 62e that is separated downward from the reference point 60 by the dimension L22 calculated by the calculation unit 24 (e.g., directly below). Next, the straight line determination unit 25 specifies a focus range 82 (diagonal line portion in FIG. 25) extending in the left-right direction in the extracted image 31, sandwiched between an approximately horizontal line 54d in the extracted image 31 that passes through the point 62d and an approximately horizontal line 54e in the extracted image 31 that passes through the point 62e. The focus range 82 includes the point 62 that is separated downward from the reference point 60 by the dimension L2, as described in the first embodiment above. Next, the straight line determination unit 25 specifies the leftmost point 44d and the rightmost point 45d on the contour line 40 that is located in the specified focus range 82. The leftmost point 44d may be determined, for example, by moving a perpendicular line in the extracted image 31 from the left side of the contour line 40 toward the contour line 40, and determining the first point that touches the contour line 40 located in the range of interest 82 as the leftmost point 44d. Alternatively, the extracted image 31 may be converted into an X-Y coordinate system, and the leftmost point 44d may be determined from the X and Y coordinates of each pixel of the contour line 40 located in the range of interest 82. The rightmost point 45d may be determined in a similar manner. Next, the straight line determination unit 25 determines a straight line 55d that passes through the point 44d and the reference point 60, and a straight line 56d that passes through the point 45d and the reference point 60. The angle acquisition unit 26 acquires the angle δ formed by the straight lines 55d and 56d. The score determination unit 27 determines the lumen size score from the angle δ acquired by the angle acquisition unit 26.

According to the fifth embodiment, the straight line determination unit 25 identifies a point 62d (second point) that is separated downward from the reference point 60 by a dimension L21 (third dimension) that is shorter than the dimension L2 (second dimension), and a point 62e (third point) that is separated downward from the reference point 60 by a dimension L22 (fourth dimension) that is longer than the dimension L2. The straight line determination unit 25 then identifies the range between the approximately horizontal line 54d in the extracted image 31 that passes through the point 62d and the approximately horizontal line 54e in the extracted image 31 that passes through the point 62e as the range of interest 82, and determines the straight line that passes through the point 44d located at the leftmost position on the contour line 40 located in the range of interest 82 and the reference point 60 as the straight line 55d. The straight line determination unit 25 also determines the straight line that passes through the point 45d located at the rightmost position on the contour line 40 located in the range of interest 82 and the reference point 60 as the straight line 56d. This determines straight lines 55d and 56d that pass through points 44d and 45d that protrude outward from the surface of the cow's abdomen and reference point 60, and the rumen size score of the cow can be determined from the angle δ formed by lines 55d and 56d, making it possible to accurately determine the rumen size score.

26 is a diagram for explaining the processing executed by the straight line determination unit 25 and the angle acquisition unit 26 in the modified example of the fifth embodiment. As shown in FIG. 26, in the modified example of the fifth embodiment, the straight line determination unit 25 specifies a point 62 that is separated downward from the reference point 60 by the dimension L2 calculated by the calculation unit 24, and specifies the range sandwiched between the approximately horizontal lines 54f and 54g in the extracted image 31 that is separated by a predetermined length above and below the point 62 as the focus range 82a (the diagonal line portion in FIG. 25). The focus range 82a includes the point 62 that is separated downward from the reference point 60 by the dimension L2. Then, the straight line determination unit 25 specifies the leftmost point 44e and the rightmost point 45e on the contour line 40 located in the focus range 82a. Next, the straight line determination unit 25 specifies a straight line 55e that passes through the point 44e and the reference point 60, and a straight line 56e that passes through the point 45e and the reference point 60. The angle acquisition unit 26 acquires the angle ε between the straight lines 55e and 56e. The score determination unit 27 determines the lumen size score from the angle ε acquired by the angle acquisition unit 26.

According to a modified example of the fifth embodiment, the straight line determination unit 25 identifies a point 62 (first point) that is located downward from the reference point 60 by the dimension L2 (second dimension), and identifies the range sandwiched between the approximately horizontal lines 54f and 54g in the extracted image 31 that are located a predetermined distance above and below the point 62 as the focus range 82a. The straight line determination unit 25 then determines the straight line passing through the reference point 60 and the leftmost point 44e on the contour line 40 located in the focus range 82a as the straight line 55e. The straight line determination unit 25 also determines the straight line passing through the reference point 60 and the rightmost point 45e on the contour line 40 located in the focus range 82a as the straight line 56e. Even in this case, the lumen size score can be accurately determined, as in the fifth embodiment.

In a modification of the fifth embodiment, the approximately horizontal lines 54f and 54g are preferably located between an approximately horizontal line in the extracted image 31 that passes through a point that is separated downward from the reference point 60 by a dimension calculated by the calculation unit 24 using a coefficient of 0.2, and an approximately horizontal line in the extracted image 31 that passes through a point that is separated downward from the reference point 60 by a dimension calculated using a coefficient of 0.4. The approximately horizontal lines 54f and 54g may be located at equal distances from the approximately horizontal line in the extracted image 31 that passes through point 62, or may be located at different distances.

As in the fifth embodiment and its modified example, the straight line determination unit 25 identifies the range between two substantially horizontal lines in the extracted image 31, including point 62 that is a distance L2 from the reference point 60, as the range of interest 82, 82a, and determines the straight lines passing through the leftmost points 44d, 44e on the contour line 40 located in the range of interest 82, 82a and the reference point 60 as straight lines 55d, 55e. In addition, the straight line determination unit 25 determines the straight lines passing through the rightmost points 45d, 45e on the contour line 40 located in the range of interest 82, 82a and the reference point 60 as straight lines 56d, 56e. In this way, the lumen size score can be determined with high accuracy.

In the fifth embodiment and its modified example, the reference point identification unit 22 may identify the reference point 60a as shown in FIG. 15 of the second embodiment, or may identify the reference point 60b as shown in FIG. 22 of the fourth embodiment. The dimension identification unit 23 may identify the dimension L3 as shown in FIG. 15 of the second embodiment, may identify the dimension L5 as shown in FIG. 18 of the third embodiment, or may identify the dimension L7 as shown in FIG. 22 of the fourth embodiment. The angle acquisition unit 26 may acquire the angle γ1 or γ2 as shown in FIG. 18 of the third embodiment, or may acquire the angle γ3 or γ4 as shown in FIG. 20 of the modified example of the third embodiment.

Sixth embodiment
In the sixth embodiment, a case will be described in which the dimension specifying unit 23 of the control unit 15 specifies the dimension indicating the size of the body in the vertical direction of the cow from the distance between the vertices of the left and right hip angles of the cow. FIG. 27 is a diagram showing the relationship between the distance between the vertices of the left and right hip angles of the cow and the cow's withers height. The horizontal axis (x-axis) of FIG. 27 is the distance between the vertices of the cow's hip angles, and corresponds to the distance between the vertex 42a of the peak 41a and the vertex 42b of the peak 41b in FIG. 5. The vertical axis (y-axis) of FIG. 27 is the cow's withers height, and corresponds to the distance L3 between the reference point 60a and the intersection point 61a in FIG. 15. Note that the horizontal and vertical axes of FIG. 27 are lengths on the drawing, and do not represent the length of the target part of the actual cow. In FIG. 27, the circles indicate each dairy cow, and the straight lines indicate approximate straight lines. As shown in FIG. 27, there is a correlation between the distance between the vertices of the cow's hip angle and the cow's withers height, and the approximate straight line is y = 1.17x + 128.73, with a coefficient of determination ^R2 of 0.623.

From the results in Figure 27, it can be seen that by determining the distance between apex 42a of peak 41a and apex 42b of peak 41b in Figure 5, it is possible to determine distance L3 in Figure 15, which is a dimension indicating the size of the cow's body in the vertical direction.

[Experiment 5]
In experiment 5, the dimension specification unit 23 of the control unit 15 determined the distance between the apex 42a of the peak 41a and the apex 42b of the peak 41b of the contour line 40 obtained by extracting the contour of the cow, and specified the distance L3 in FIG. 15 as the dimension indicating the size of the body in the vertical direction of the cow from this distance and the approximation line in FIG. 27. Then, as shown in FIG. 16, the line determination unit 25 determined the lines 55a and 56a using the dimension L3 specified by the dimension specification unit 23, and the angle acquisition unit 26 acquired the angle β formed by the lines 55a and 56a. FIG. 28 shows the results of an experiment investigating the correlation between the rumen size score and the angle β acquired by the angle acquisition unit 26. The horizontal axis (x-axis) of FIG. 28 is the rumen size score specified by an expert using a judgment sheet, and the vertical axis (y-axis) is the angle β acquired by the angle acquisition unit 26. Each point of ○, △, and □ in FIG. 28 indicates each dairy cow in the experiment, and the straight line indicates an approximation line. Moreover, ◯, △, and □ indicate cases where the coefficients used by the calculation unit 24 to calculate the dimension L4 are different from one another. The coefficients used to calculate the dimension L4 are ◯: 0.3, △: 0.35, and □: 0.4.

As shown in Fig. 28, even when the dimension L3 indicating the size of the cow's body in the vertical direction was determined from the distance between the apex 42a of the crest 41a and the apex 42b of the crest 41b of the contour line 40, the larger the rumen size score, the larger the angle β, and the result showed a correlation between the rumen size score and the angle β. The approximate line for the results of ○ was y = 5.88x + 54.65, and the coefficient of determination ^R2 was 0.754. The approximate line for the results of △ was y = 7.23x + 43.00, and the coefficient of determination ^R2 was 0.687. The approximate line for the results of □ was y = 5.15x + 39.61, and the coefficient of determination ^R2 was 0.439.

According to the sixth embodiment, the dimension determination unit 23 determines the dimension indicating the vertical body size of the cow from the distance between the vertices 42a, 42b of the two peaks 41a, 41b on both the left and right sides of the upper part of the contour line 40. Even in this case, the rumen size score can be determined with high accuracy, as described in FIG. 28.

In the sixth embodiment, the worker may operate the input unit 11 to specify the vertices of the left and right hip angles of the cow, and the dimension specification unit 23 may specify a dimension indicating the size of the cow's body in the vertical direction from the distance between the vertices of the left and right hip angles specified by the worker.

In the above embodiments, the evaluation device that determines the rumen size score of the cow is a portable information device that is carried by the worker and has an image capturing unit 13. However, other cases are also possible. For example, a stationary information device such as a server that can communicate with other devices connected to a network may be used. In this case, the worker transmits the image of the cow captured using a portable information device such as a smartphone to a stationary information device such as a server via a network. The stationary information device performs the same process as the evaluation device 100 to determine the rumen size score. Then, the stationary information device transmits the determined rumen size score to a portable information device carried by the worker via a network. This reduces the load on the smartphone. In addition, a camera may be installed in the cowshed so that it can be fixed or moved, and the portable information device such as a smartphone and the stationary information device such as a server may acquire an image captured from approximately directly behind the cow in a standing position by the camera installed in the cowshed.

Note that the above experiments 1 to 5 were conducted on dairy cows, but it is believed that similar results would be obtained for beef cows, which have similar skeletal structures. Therefore, the above embodiments can be applied not only to dairy cows but also to beef cows.

Although the embodiment of the present invention has been described in detail above, the present invention is not limited to such a specific embodiment, and various modifications and variations are possible within the scope of the gist of the present invention as described in the claims.

REFERENCE SIGNS LIST 10 Display unit 11 Input unit 12 Communication unit 13 Imaging unit 14 Memory unit 15 Control unit 20 Image acquisition unit 21 Extraction unit 22 Reference point identification unit 23 Dimension identification unit 24 Calculation unit 25 Straight line determination unit 26 Angle acquisition unit 27 Score identification unit 30 Captured image 31 Extracted image 40 Contour line 41a, 41b Mountain portion 42a, 42b Vertex 43a, 43b Tip 44, 44a, 44b, 44c, 44d, 44e Points 45, 45a, 45b, 45c, 45d, 45e Points 50 Line segment 51, 52 Inclined straight line 53, 53a, 53b Straight line 54, 54a, 54b, 54c, 54d, 54e, 54f, 54g Approximately horizontal line 55, 55a, 55b, 55c, 55d, 55e Straight line 56, 56a, 56b, 56c, 56d, 56e Straight line 57, 57a Reference line 59 Perpendicular line 60, 60a, 60b Reference point 61, 61a, 61b Intersection point 62, 62a, 62b, 62c, 62d, 62e Point 63a, 63b Point 70 Storage device 71 Memory 72 Processor 73 Communication interface 74 Input device 75 Display device 76 Imaging device 77 Bus 78 Cow evaluation program 80 Area near the left-right center and near the upper end 81a, 81b Heel area 82, 82a: range of interest 90: abdomen 91: buttocks 93, 94: points 95, 96: straight line 100: evaluation device

Claims (14)

an image acquisition unit for acquiring a two-dimensional image of a cow in a standing position from substantially directly behind;
an extraction unit that extracts a contour line of the cow from the captured image;
A size specifying unit that specifies a first size indicating a body size of the cow in the up-down direction from the captured image by a predetermined method;
a straight line determination unit that identifies, by a predetermined method, at least one of two points on the contour line located on substantially both the left and right sides of a first point that is spaced downward from a reference point that is near the left-right center and near the upper end of the cow by a second dimension that is shorter than the first dimension, and determines at least one of a first straight line that passes through one of the two points and the reference point, and a second straight line that passes through the other point and the reference point;
an angle acquisition unit that acquires an angle between a reference line specified from the captured image and the first straight line or the second straight line, or an angle between the first straight line and the second straight line;
A cow evaluation device comprising: a score determination unit that determines the rumen size score of the cow from the angle acquired by the angle acquisition unit. The cattle evaluation device according to claim 1, wherein the straight line determination unit identifies the first point and identifies as the two points two intersections where an approximately horizontal line in the captured image passing through the first point intersects with the contour line. The cattle evaluation device according to claim 1, wherein the straight line determination unit determines a range between two substantially horizontal lines in the captured image, including the first point, as a range of interest, and determines the leftmost point located on the left side and the rightmost point located on the right side of the contour line located in the range of interest as the two points. The cattle evaluation device according to claim 3, wherein the straight line determination unit identifies a second point that is separated downward from the reference point by a third dimension that is shorter than the second dimension, and a third point that is separated downward from the reference point by a fourth dimension that is longer than the second dimension, and identifies the range of interest as being between an approximately horizontal line in the captured image that passes through the second point and an approximately horizontal line in the captured image that passes through the third point. The cattle evaluation device according to claim 3, wherein the straight line determination unit identifies the first point and identifies the range of interest as a range sandwiched between two substantially horizontal lines in the captured image that are a predetermined distance above and below the first point. a reference point specifying unit that specifies, as the reference point, an approximate midpoint of a line segment connecting vertices of two peaks present on both the left and right sides of an upper portion of the contour line or a topmost point on the uppermost side of the contour line;
A cow evaluation device as described in any one of claims 1 to 5, wherein the dimension identification unit identifies as the first dimension the distance between the reference point and an approximately horizontal line in the captured image passing through one of two tips located below the contour line, or a straight line passing through the two tips. The cattle evaluation device according to claim 6, wherein, when the reference point identification unit identifies the approximate midpoint as the reference point, the second dimension is a value obtained by multiplying the first dimension by a coefficient of 0.2 to 0.4, and when the reference point identification unit identifies the top point as the reference point, the second dimension is a value obtained by multiplying the first dimension by a coefficient of 0.3 to 0.4. The cattle evaluation device according to any one of claims 1 to 5, wherein the dimension determination unit determines the first dimension from the distance between the apexes of two peaks present on both the left and right sides of the upper part of the contour line. The cattle evaluation device according to any one of claims 6 to 8, wherein the two peaks correspond to the hip angles of the cattle. The cow evaluation device according to claim 6 or 7, wherein the highest point is a part corresponding to the base of the cow's tail. The cattle evaluation device according to any one of claims 1 to 5, wherein the dimension specification unit specifies the distance between the reference point and an approximately horizontal line in the captured image passing through one of the heels of the two hind legs of the cow, a straight line passing through both heels of the two hind legs of the cow, an approximately horizontal line in the captured image passing through one of the hooves of the two hind legs of the cow, or a straight line passing through both hooves of the two hind legs of the cow as the first dimension. The cattle evaluation device according to claim 11, wherein when the dimension specification unit specifies the distance between the reference point near the anus of the cattle and the approximately horizontal line passing through one heel or the straight line passing through both heels as the first dimension, the second dimension is a value obtained by multiplying the first dimension by a coefficient of 0.25 to 0.55. The cattle evaluation device according to any one of claims 1 to 12, wherein the reference line is an approximately vertical line or an approximately horizontal line in the captured image. A two-dimensional image of a cow standing upright is captured from directly behind.
Extracting a contour line of the cow from the captured image;
Identifying a first dimension indicating a body size of the cow in a vertical direction from the captured image by a predetermined method;
Identifying, by a predetermined method, at least one of two points on the contour line located on substantially both the left and right sides of a first point that is spaced downward from a reference point that is near the center in the left-right direction and near the upper end of the cow by a second dimension that is shorter than the first dimension, and determining at least one of a first straight line passing through one of the two points and the reference point, and a second straight line passing through the other point and the reference point;
acquiring an angle between a reference line identified from the captured image and the first straight line or the second straight line, or an angle between the first straight line and the second straight line;
determining a rumen size score for the cow from the acquired angle;
A method for evaluating cattle, characterized in that processing is carried out by a computer.

Images