JP7626366B2 - Denseness estimation device, denseness estimation method, and program - Google Patents

Description

The present invention relates to a congestion estimation device, a congestion estimation method, and a program.

Conventionally, methods for controlling the rearing environment to efficiently grow livestock such as chickens have been developed. For example, Patent Document 1 discloses a chicken coop ventilation system that controls the ventilation of the chicken coop, the operation of the heat exchanger, etc., based on measurements from an outside air thermometer that measures the air temperature outside the chicken coop, an inside thermometer that measures the air temperature inside the chicken coop, etc., as a method for controlling the temperature inside the chicken coop.

JP 2019-10010 A

In the control method of Patent Document 1, in order to control the rearing environment for the chickens, the air temperature is measured using a thermometer or the like. In addition, carbon dioxide is measured and ventilation control is performed to realize a more suitable rearing environment.

However, when livestock such as chickens are raised in a breeding house, the animals tend to crowd together in areas where the air temperature and floor (ground) temperature are comfortable, which can lead to higher carbon dioxide concentrations and other problems that can worsen the environment for their growth. In this way, the density of animals in a breeding house affects the environment in which the animals are raised and, ultimately, the rate at which the animals grow, but it is difficult to accurately estimate the density of animals in real time using measurements such as thermometers and carbon dioxide concentration measuring devices.

The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide a density estimation device, a density estimation method, and a program that can instantly and accurately estimate animal density by estimating the density of animals based on images of the inside of an animal enclosure.

In order to achieve the above object, a congestion estimation device according to a first aspect of the present invention comprises:
an image data processing unit that generates an estimation image from an inside image, which is an image of an animal kept in the animal house;
a density estimation unit that estimates the density of the animals based on the estimation image,
The density estimation unit estimates the density of the animals by performing kernel density estimation based on the estimation image.

A storage unit that stores the estimation result of the animal density.
A display unit that displays the estimation result,
The congestion estimation unit is
calculating a feature amount for evaluating the density from the density calculated by performing the kernel density estimation;
storing the estimation result based on the feature amount in the storage unit and displaying it on the display unit;
This may also be the case.

The image data processing unit further comprises:
generating the binarized estimation image based on the house image, and detecting pixels in which the animal is photographed from the estimation image;
The density estimation unit performs the kernel density estimation based on coordinate information of the detected pixels.
This may also be the case.

Also, the kernel function of the kernel density estimation is a normal distribution.
This may also be the case.

Also, a camera for capturing images of the inside of the building;
A camera control unit that controls the camera,
The camera control unit captures the interior image at a predetermined time interval.
This may also be the case.

Moreover, the congestion estimation unit
calculating an average value of the density of each pixel calculated from the estimation image as the feature amount at the estimation time;
This may also be the case.

In addition, in a congestion estimation method according to a second aspect of the present invention,
An estimation image is generated from an inside image, which is an image of an animal kept in the animal house;
The density of the animals is estimated by performing kernel density estimation based on the estimation image.

Moreover, a program according to a third aspect of the present invention comprises:
Computer,
an image data processing unit that generates an estimation image from an inside image, which is an image of an animal kept in the animal house;
a density estimation unit that estimates the density of the animals by performing kernel density estimation based on the estimation image;
Function as.

The animal density estimation device, animal density estimation method, and program of the present invention estimate animal density by kernel density estimation based on images of the inside of the animal enclosure, making it possible to instantly and accurately estimate animal density.

1 is a block diagram showing a configuration of a congestion estimation device according to an embodiment of the present invention. 1 is a flowchart showing a flow of congestion estimation according to an embodiment. 13A and 13B are diagrams showing examples of images related to density estimation, in which (A) is a building interior image _Ih , (B) is a grayscale image _Ig , and (C) is an estimation image _Ib . 1A is a conceptual diagram showing a method for calculating density, in which (A) is an example of an estimation image _Ib , (B) is an example of coordinate information φ stored in a storage unit, and (C) is an example of density S for each pixel. 13 is an example of an image _Ih of inside a building at a specific time in winter. 13A is an example of a distribution map of density S corresponding to each pixel of the building interior image _Ih , and FIG. 13B is a histogram showing the frequency of each density in FIG. 6 is a graph showing a change over time in a feature amount F that indicates a density estimated from the building interior image _Ih in FIG. 5 . 8A and 8B show data at times when the density is low in FIG. 7, where (A) is the building interior image _Ih , (B) is the estimation image _Ib , (C) is the density S at each coordinate, and (D) is a histogram of the density. 8A and 8B show data at times when the density was high in FIG. 7, where (A) is the building interior image _Ih , (B) is the estimation image _Ib , (C) is the density S at each coordinate, and (D) is a histogram of the density. 11 is a graph showing an example of a change over time in a feature amount F that indicates a density in summer. 11A and 11B show data at times when the density was high in FIG. 10, where (A) is the building interior image _Ih , (B) is the estimation image _Ib , (C) is the density S at each coordinate, and (D) is a histogram of the density. 11A and 11B show data at times when the density is low in FIG. 10, where (A) is the building interior image _Ih , (B) is the estimation image _Ib , (C) is the density S at each coordinate, and (D) is a histogram of the density.

The following describes a population density estimation device 1 according to an embodiment of the present invention with reference to the drawings. In this embodiment, the population density estimation device 1 that estimates the population density of chickens raised in a breeding house is used as an example.

As shown in the block diagram of FIG. 1, the crowding estimation device 1 includes a camera 10 and a control unit 20.

The camera 10 is a camera that captures an image of a target area in a breeding house for estimating chicken density. The camera 10 transmits data of an image _Ih of the captured target area inside the house to the control unit 20. The type, resolution, and other characteristics of the camera 10 are not particularly limited, but it is preferable that the camera has a resolution that allows animals in the target area to be individually recognized.

The control unit 20 is, for example, a computer device, and as shown in FIG. 1, includes a control unit 21, a memory unit 22, a display unit 23, and an input unit 24.

The control unit 21 is composed of a central processing unit (CPU), a read only memory (ROM), a random access memory (RAM), a crystal oscillator, etc., and controls the operation of the camera 10. The control unit 21 also estimates the density of animals based on the inside-house image _Ih captured by the camera 10. The control unit 21 loads various operation programs and data stored in the ROM, storage unit 22, etc. of the control unit 21 into the RAM and operates the CPU, thereby realizing each function of the control unit 21 shown in FIG. 1. In this way, the control unit 21 operates as a camera control unit 211 and an estimation unit 212.

The camera control section 211 controls the camera 10 connected to the control unit 20 to capture an image of the target area and obtain an interior image _Ih .

The estimation unit 212 includes an image data processing unit 212-1 and a density estimation unit 212-2. The image data processing unit 212-1 performs image processing based on the acquired inside-the-house image _Ih to generate an estimation image _Ib for calculating the density of chickens in the target area. The estimation image _Ib is an image for extracting basic data used in calculating the density. The estimation image _Ib according to this embodiment is a binary image generated based on the inside-the-house image _Ih . The image data processing unit 212-1 also extracts coordinate data of pixels at positions where chickens are present from the estimation image _Ib , and uses this as basic data for calculating the density.

The density estimation unit 212-2 performs kernel density estimation based on the estimation image _Ib generated by the image data processing unit 212-1, and estimates the density by extracting the feature amount F. The detailed estimation procedure will be described later.

The storage unit 22 is a non-volatile memory such as a hard disk or a flash memory, and stores an algorithm for estimating chicken density from the inside-house image _Ih , information on the estimated density, and the like.

The display unit 23 is a display device, such as a liquid crystal monitor, provided in the control unit 20, which is a computer device. The display unit 23 displays the inside-building image I _h captured by the camera 10, information related to the density estimated by the estimation unit 212, and the like.

The input unit 24 is an input device for inputting parameters related to shooting, such as the angle of view of the camera 10, the time interval for calculating the density, and instructions to start and end density estimation. The input unit 24 is a keyboard, touch panel, mouse, etc., provided in the control unit 20.

The control unit 20 also communicates with the camera 10 via a communication interface (communication I/F). Specifically, the camera control unit 211 transmits control information to the camera 10 and acquires image data of the captured inside-house image _Ih . The control unit 21 also stores information such as the estimated chicken density and the estimated time of density in the memory unit 22.

The following describes in detail the congestion estimation method according to this embodiment with reference to the flowchart in Figure 2.

When the crowding estimation is started by inputting an estimation start command to the input section 24, the camera control section 211 of the control unit 20 controls the camera 10 to capture an image of a target area in the breeding house at a predetermined estimation time, and obtains an inside-house image _Ih (step S11). Fig. 3(A) is an example of the captured inside-house image _Ih .

In this embodiment, a predetermined area in the breeding house is photographed using one camera 10, but the present invention is not limited to this, and multiple cameras 10 may be used to obtain an image _Ih of the inside of the breeding house of the target area corresponding to each camera 10. This makes it possible to estimate the density of animals over a wider range in the breeding house. In addition, the same target area may be photographed from different directions by multiple cameras 10, which makes it possible to perform estimation with less error.

The camera control unit 211 stores data of the inside building image _Ih captured by the camera 10 in the storage unit 22 and transmits the data to the image data processing unit 212-1 (step S12). The inside building image _Ih may be a color image or a grayscale image. The inside building image _Ih in this embodiment is a color image.

The image data processing unit 212-1 converts the indoor image _Ih into a grayscale image _Ig (FIG. 3B) based on the data received in step S12 (step S13). The method of converting the indoor image Ih into a grayscale image Ig (FIG. 3B) is not particularly limited, and any known method such as averaging RGB or using gamma correction can be used.

Image data processing unit 212-1 binarizes grayscale image _Ig generated in step S13 to generate estimation image _Ib (step S14). The threshold value for the binarization process is set in advance to a value between the luminance of the area of the chicken in the image to be processed and the luminance of the floor of the breeding house, which is the background. As a result, as shown in the example of FIG. 3(C), estimation image _Ib is generated as a binary image in which the chicken area is white and the background area other than the chicken is black. That is, the binarization process detects the pixels in which the chicken is photographed as white pixels.

Image data processing unit 212-1 stores coordinate information φ=[φ ₁ , φ ₂ ]=[x, y] of white _pixels in estimation image Ib in storage unit 22 (step S15).

The density estimation unit 212-2 then calculates the density S based on the coordinate information φ of the white pixels stored in the storage unit 22 (step S16). The density estimation unit 212-2 also calculates a feature value F from the calculated density S, and displays and stores it as an estimation result of the chicken density (step S17). Specifically, the density estimation unit 212-2 calculates the density S using kernel density estimation. In this embodiment, kernel density estimation is performed using a kernel function of normal distribution to calculate the density S(φ(k), φ), as shown in the following formula.

Here, N is the number of white pixel data stored in the storage unit, φ _i (k) is the i-th element of the coordinate information, and h _i is the bandwidth of the i-th element of the coordinate information.

4A to 4C are examples of calculating density S for an image of 3×3 pixels. Fig. 4A is a diagram showing the distribution of white pixels and black pixels, and Fig. 4B is a diagram showing coordinate information φ stored in storage unit 22. Note that in this example, for the purpose of explanation, each pixel is marked with a symbol such as a circle or a star. The parameters of the above formula in this example are as follows:
k=1, 2,..., 9
N=4 (number of white pixel data)
φ(k): Coordinate information of the k-th pixel φ: Coordinate information of all white pixels stored in the memory unit 22 φ ₁ (k): x-coordinate of the k-th pixel φ ₂ (k): y-coordinate of the k-th pixel h ₁ : Bandwidth on the x-axis h ₂ : Bandwidth on the y-axis

Figure 4 (C) is a diagram showing the density S of each pixel calculated for Figure 4 (A). As shown in Figure 4 (C), φ(5) (the pixel indicated by a star in the figure) is located in the center of all the white pixels, so it can be seen that the density S(φ(5), φ) of φ(5) is high.

Here, the bandwidth h _i is a parameter that can be set by any method, and can be determined by a known method such as a method using standard deviation. In this embodiment, the bandwidth h is set by the user based on the distribution of the calculated density S, and is appropriately selected so that the density of animals in the target area can be easily determined. More specifically, the user sets the bandwidth h (= h ₁ ) for the x-axis and the bandwidth h (= h ₂ ) for the y-axis while checking the distribution of the calculated density S.

Fig. 5 is an example of an interior image _Ih at a specific time between 1:00 and 14:00 on a certain day in winter. Fig. 6(A) is a distribution diagram of the density S corresponding to each pixel of the interior image _Ih at a certain time, calculated by the density estimation method according to the present embodiment. Fig. 6(B) is a histogram showing the frequency of each density.

In this embodiment, the density estimation unit 212-2 calculates the average density S of each pixel, and the calculated average value is used as the feature F that represents the density of chickens at that time.

In this embodiment, in order to simplify the calculation process for calculating the feature value F, the density S corresponding to half of the pixels is extracted from the density S corresponding to all the pixels in the interior image _Ih , and the average value is calculated. This makes it possible to calculate the feature value F with a smaller calculation load. In this embodiment, the average value of the density S is used as the feature value F, but this is not limited to this, and the median, standard deviation, a combination of these with the average value, etc. may also be used.

Fig. 7 is a graph showing the change over time in feature value F, which represents the chicken density estimated from the inside-house image _Ih in Fig. 5. A comparison of the inside-house image _Ih in Fig. 5 with the graph in Fig. 7 shows that the density estimation method according to the present embodiment is able to appropriately estimate the chicken density in the breeding house.

More detailed estimation data is shown in Figures 8(A)-(D) and 9(A)-(D). Figures 8(A)-(D) show the building interior image _Ih (Figure 8(A)), the estimation image _Ib (Figure 8(B)), which is a binary image, the density S of each coordinate (Figure 8(C)), and the histogram of density S (Figure 8(D)). Figures 9(A)-(D) show the building interior image _Ih (Figure 9(A)), the estimation image _Ib (Figure 9(B)), the density S of each coordinate (Figure 9(C)), and the histogram of density S (Figure 9(D)) at the time when the density is high (11:00).

From Figures 8 (A)-(D) and Figures 9 (A)-(D), we can see that the times when chicken density is low, i.e., when chickens are sparse and a lot of ground is visible, and the times when chicken density is high, i.e., when chickens are densely packed in the target area and little ground is visible, can be appropriately estimated.

The above Fig. 7, Fig. 8 (A)-(D), and Fig. 9 (A)-(D) are examples of data for one day in winter (December). On the other hand, Figs. 10-12 show examples of data for one day in summer (July). Fig. 10 is a graph showing changes in the feature value F that represents the density. Figs. 11 (A)-(D) are an image _Ih (Fig. 11 (A)) of the inside of a building at a time of high density (3 o'clock), an estimation image _Ib (Fig. 11 (B)), density S of each coordinate (Fig. 11 (C)), and a histogram of density S (Fig. 11 (D)). 12(A) to (D) show the house interior image _Ih (FIG. 12(A)), the estimation image _Ib (FIG. 12(B)), the density S at each coordinate (FIG. 12(C)), and a histogram of the density S (FIG. 12(D)) at a time when chicken density is low (12 o'clock). It can be seen from FIGS. 10 to 12 that the density estimation method according to the present embodiment is able to appropriately estimate times when chicken density is low and times when chicken density is high.

Returning to the flowchart of FIG. 2, the control unit 21 stores data such as the density S calculated by the density estimation unit 212-2 and the feature quantity F as an estimation result in the memory unit 22, and also displays the data on the display unit 23 (step S18).

The control unit 21 repeats the above-described density estimation until the density estimation is terminated due to an input of an end of estimation to the input unit 24, the end of a preset estimation period, etc. (NO in step S19). The control unit 21 terminates the process when a termination condition is satisfied (YES in step S19), such as an input of an end of estimation to the input unit 24, the end of a preset estimation period, etc.

As described above, in the animal density estimation device 1 according to the present embodiment, the animal density is estimated by kernel density estimation based on the inside-house image _Ih obtained by photographing the animals in the house, and therefore the animal density can be estimated immediately. Furthermore, since the animal density is estimated based on the inside-house image _Ih , it is possible to estimate the animal density more accurately compared to a method of estimating the animal density by placing measuring devices such as a thermometer and a carbon dioxide concentration meter in the house.

In this embodiment, the density is estimated directly using the inside-house image _Ih captured inside the breeding house, but this is not limiting. For example, the density may be estimated after processing such as distortion correction and trimming of the inside-house image _Ih . This allows for more accurate estimation of the density. In this case, an image correction process may be inserted between steps S12 and S13 in the flowchart of FIG. 2, and the corrected image may be used to perform grayscale conversion and subsequent processes.

In addition, in this embodiment, the density of chickens raised in a chicken coop is estimated, but the present invention is not limited to this, and the density of other animals such as pigs, rabbits, ducks, etc. may also be estimated. In this case, it is preferable to adjust the background color, for example by painting the inside of the coop, so that the difference in brightness in the image data between the animals and the background, such as the ground of the coop, becomes larger.

In the present embodiment, the camera 10 and the control unit 20 are directly connected, but this is not limited thereto. For example, the camera 10 and the control unit 20 may be connected via a network. This allows the density of animals in a breeding facility to be estimated from a remote location, so that even in cases where there are multiple breeding facilities or a large breeding facility, the information can be aggregated in one place and the environment in which the animals are raised can be easily evaluated.

In the present embodiment, the population density estimation device 1 includes the camera 10, and estimates the population density instantly based on the house image _Ih captured by the camera 10, but is not limited to this. For example, the population density may be estimated based on past house images _Ih stored in the storage unit 22. This makes it possible to read out necessary data from the house images _Ih stored in the server, estimate the animal population density, and use the data for evaluating the relationship between the rearing environment of the animals and the growth rate.

The animal density estimation method according to the above embodiment can be realized using a normal computer system. For example, a computer program for executing the density estimation according to the above embodiment can be stored in a computer-readable recording medium such as a USB memory or a DVD-ROM and distributed, and the computer program can be installed on a computer to cause the computer to function as the control unit 20 of the density estimation device 1 that executes the above density estimation.

The present invention is suitable for a rearing environment evaluation system that estimates the density of animals kept in a breeding house and evaluates the rearing environment. In particular, the present invention is suitable for a rearing environment evaluation system that estimates the density of chickens in a chicken house where a large number of animals are kept.

1 Denseness estimation device, 10 Camera, 20 Control unit, 21 Control unit, 211 Camera control unit, 212 Estimation unit, 212-1 Image data processing unit, 212-2 Denseness estimation unit, 22 Storage unit, 23 Display unit, 24 Input unit

Claims (6)

an image data processing unit that generates an estimation image from an inside image, which is an image of an animal kept in the animal house;
a density estimation unit that estimates the density of the animals based on the estimation image,
The image data processing unit includes:
generating the binarized estimation image based on the house image, and detecting pixels in which the animal is photographed from the estimation image;
The congestion estimation unit is
calculating a density of the animals at each pixel by performing kernel density estimation for each pixel of the estimation image based on coordinate information of the pixel detected by the image data processing unit;
estimating the density of animals by averaging the density of animals for each pixel and calculating a feature amount for evaluating the density of animals in the estimation image ;
A congestion estimation device comprising: A storage unit that stores the estimation result of the animal density;
A display unit that displays the estimation result,
The congestion estimation unit is
storing the estimation result based on the feature amount in the storage unit and displaying it on the display unit;
The congestion estimation device according to claim 1 . the kernel function of the kernel density estimation is a normal distribution;
3. The congestion estimation device according to claim 1 or 2 . A camera for capturing images of the interior of the building;
A camera control unit that controls the camera,
The camera control unit captures the interior image at a predetermined time interval.
The congestion estimation device according to any one of claims 1 to 3 . A binarized estimation image is generated based on an interior image, which is an image of an animal kept in the animal house;
Detecting pixels in which the animal is photographed from the estimation image;
calculating a density of the animal at each pixel by performing kernel density estimation for each pixel of the estimation image based on coordinate information of the detected pixel ;
estimating the density of animals by averaging the density of animals for each pixel and calculating a feature amount for evaluating the density of animals in the estimation image ;
A method for estimating congestion, comprising: Computer,
an image data processing unit that generates a binarized estimation image based on an interior image, which is an image of an animal kept in a breeding house , and detects pixels in which the animal is photographed from the estimation image ;
a density estimation unit that estimates the animal density by calculating the animal density of each pixel by performing kernel density estimation for each pixel of the estimation image based on coordinate information of the pixel detected by the image data processing unit , and averaging the animal densities of each pixel to calculate a feature amount for evaluating the animal density in the estimation image ;
A program that functions as a