JP7101342B2 - Dead bird estimation device, dead bird estimation method, program and recording medium - Google Patents

Description

The present invention relates to a dead bird estimation device, a dead bird estimation method, a program and a recording medium.

Cages are usually used to raise chickens in poultry houses. In the poultry house, cage lanes in which a plurality of cages are continuous are generally arranged, and about 10 chickens are housed in each cage (about 4,200 cm ² ). In the cage lane, the cages are laminated so that, for example, 80 to 140 cages are sewn over a length of about 70 to 80 m in the planar direction, and for example, about 2 to 4 stages are formed in the vertical direction. Has been done.

In the case of chickens for meat, chickens that have died in the poultry house can be prevented from being distributed to the market by discarding them when they are collected from the poultry house. However, with regard to eggs, there are cases where old eggs are mixed with dead birds. The cage has an inclined surface in which the inner floor is lowered in the front direction, and an egg collecting lane is provided in front of the inclined surface. Therefore, the eggs laid by the live birds in the cage roll on the inclined surface and are automatically collected in the egg collection lane. However, as mentioned above, when an egg is caught by a dead bird and becomes old and then moves to the egg collection lane, the appearance of the egg indicates whether it is a fresh egg or an old egg. It is difficult to judge. For this reason, dead birds in cages need to be removed quickly to prevent contamination with old eggs.

However, in the actual field, the current situation is that workers visually check each cage to determine whether or not there are dead birds. As described above, in the poultry house, a plurality of cages are arranged over a long distance, and further, a plurality of cages are arranged in the height direction. Therefore, it takes a lot of labor for the worker to visually check each one. This confirmation is performed every day before feeding and in the dark, but it takes 2 hours even if 4 people perform it per poultry house of a general size. As described above, it takes time to confirm all cages in the poultry house, so frequent confirmation is practically difficult. And even in a harsh environment, long hours of work can lead to oversight of dead birds. In addition, if dead birds are left in their cages for extended periods of time, they may be trampled by live birds in the same cage. As mentioned above, since it is difficult to confirm frequently, if the dead bird is trampled by the next confirmation, there is also a problem that the dead bird cannot be confirmed as a result.

On the other hand, from the viewpoint of solving the labor shortage, it has been reported that the cage of the poultry house is monitored by a surveillance camera (Patent Document 1). However, this method only photographs the cages with a camera, and as a result, the worker has no choice but to check the captured contents on the monitor and determine the presence or absence of dead birds in each cage.

Special fair No. 7-85696

Therefore, an object of the present invention is to provide, for example, a new system capable of easily estimating the presence or absence of dead birds in a cage (hereinafter referred to as dead bird estimation) in real time.

In order to achieve the above object, the dead bird estimation device of the present invention is used.
Includes image acquisition section, cage extraction section, and dead bird estimation section
The image acquisition unit
Acquire multiple scan images of scanned cage lanes in which cages are continuously arranged,
The cage extraction unit
From the plurality of scanned images, cage images for each of the continuously arranged cages are extracted.
The dead bird estimation unit
For each of the extracted cage images, it is estimated whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation information for estimating whether or not the bird in the cage in the image is a dead bird.
It is characterized by that.

The dead bird estimation method of the present invention
Including image acquisition process, cage extraction process, and dead bird estimation process
The image acquisition step is
Acquire multiple scan images of scanned cage lanes in which cages are continuously arranged,
The cage extraction step is
From the plurality of scanned images, cage images for each of the continuously arranged cages are extracted.
The dead bird estimation process is
For each of the extracted cage images, it is estimated whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation information for estimating whether or not the bird in the cage in the image is a dead bird.
It is characterized by that.

The program of the present invention is characterized in that a computer executes the dead bird estimation method of the present invention.

The recording medium of the present invention can be read by a computer on which the program of the present invention is recorded.

According to the present invention, it is possible to easily estimate dead birds in a cage in real time.

FIG. 1 is a block diagram showing an example of the dead bird estimation device of the first embodiment. FIG. 2 is a schematic view showing an example of a cage lane in a poultry house. FIG. 3 is an image showing the state of dead and live birds in the cage. FIG. 4 is a block diagram showing an example of the hardware configuration of the dead bird estimation device according to the first embodiment. FIG. 5 is a block diagram showing another example of the hardware configuration of the dead bird estimation device according to the first embodiment. FIG. 6 is a flowchart showing an example of the dead bird estimation method of the first embodiment. FIG. 7 is a block diagram showing an example of the dead bird estimation device of the second embodiment. FIG. 8 is a schematic view showing an example of the dead bird estimation device of the third embodiment. FIG. 9 is a schematic view showing the relationship between the angle between the cage and the camera in the third embodiment.

In the present invention, the type of bird that is the target of estimation of dead birds is, for example, a bird that is bred in a cage for eggs, and specific examples thereof include an egg-collecting chicken (chicken).

According to the present invention, for example, the presence of dead birds in each cage can be estimated in real time by performing image processing while photographing the cage lane. Therefore, for example, the labor and working time of the worker can be reduced, and the occurrence of dead birds in the cage lane can be managed every day, several times a day, or continuously as needed.

An embodiment of the present invention will be described with reference to the drawings. The present invention is not limited to the following embodiments. In each of the following figures, the same parts are designated by the same reference numerals. Further, the explanations of the respective embodiments can be referred to each other's explanations unless otherwise specified. Further, the configurations of the embodiments can be combined unless otherwise specified. In the following embodiment, a chicken will be described as an example of a dead bird estimation target, but as described above, the present invention is not limited to this example.

[Embodiment 1]
An example of the dead bird estimation device and the dead bird estimation method of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an example of the dead bird estimation device of the present embodiment. The dead bird estimation device 1 includes an image acquisition unit 11, a cage extraction unit 12, and a dead bird estimation unit 13. The dead bird estimation device 1 further includes, for example, a storage unit 10, and the storage unit 10 further includes a dead bird estimation information storage unit 101. The dead bird estimation device 1 further includes, for example, an output unit 14. The dead bird estimation device 1 is also referred to as, for example, a dead bird estimation system. The dead bird estimation device 1 may be, for example, one dead bird estimation device including the above-mentioned parts, or may be a dead bird estimation device in which each part can be connected via a communication network. The communication line network is not particularly limited, and a known communication line network can be used, and may be wired or wireless. Specifically, for example, an internet line, a telephone line, a LAN (Local Area Network), WiFi (Wireless Fidelity). ) Etc.

In the storage unit 10, the dead bird estimation information storage unit 101 stores information for estimating whether or not the bird in the cage in the image is a dead bird as the dead bird estimation information. The type of the information is not particularly limited, and examples thereof include the following.

The first example of the above information is a dead bird estimation model. The dead bird estimation model is a model that represents the relationship (algorithm) between the image and the state of the bird (dead bird or live bird) in the image. The dead bird estimation model is generated by training using, for example, a dead bird image of a dead bird in a cage and a live bird image of a live bird in the cage as training data. The learning includes so-called machine learning. In this case, the dead bird estimation device 1 may further include, for example, a learning unit that generates the dead bird estimation model. For machine learning using the image as learning data, for example, a commercially available system can be used, and as a specific example, FieldAnalyst (product name, NEC Solution Innovators, Inc.) or the like can be used (hereinafter, the same applies). The generated dead bird estimation model is stored in the dead bird estimation information storage unit 101.

The second example of the above information is the dead bird image information showing the state of the dead bird in the cage and the live bird image information showing the state of the live bird in the cage. The dead bird image information and the live bird image information may be, for example, the image itself or the information obtained from the image. In the latter case, for example, information on the characteristic points of the dead bird in the dead bird image and information on the characteristic points of the live bird in the live bird image can be mentioned, for example, the dead bird image and the live bird image. It can be obtained by performing image analysis and extracting each feature point and its feature amount. The method of image analysis is not particularly limited, and for example, a known method can be used.

The dead bird image and the live bird image are preferably images including, for example, a lower region of the cage (that is, a region including the legs of the live bird when it is present). Specifically, the lower region is a region including, for example, below the knee (ie, the yoke and the leg) when the chicken in the cage is a live bird, preferably below the heel (ie, the leg). ) Is included. During the active time of the chicken, the live bird is usually standing, so the leg of the chicken should be visible in the lower area, but in the case of the dead bird, the leg is confirmed because it is in a crouched state. It's hard to do. Therefore, by using an image including the lower region of the cage, more accurate estimation of dead birds becomes possible. Further, the dead bird image and the live bird image are preferably images taken during the chicken activity time zone, for example, for the same reason. During the time when the chicken is resting, it is difficult to determine whether the chicken is crouching or dead. Therefore, by using the image of the active time zone, it is possible to estimate the dead bird more accurately. The activity time zone is not particularly limited, and is, for example, a time zone during feeding (around 4:00 am) and the like.

Here, FIG. 2 shows an image of the lower region of the cage during the chicken activity time zone. FIG. 2A is an image of the lower region of the cage and an image of a dead bird, and FIG. 2B is an image of the lower region of the cage and only a live bird is shown. It is an image. Both images are images showing the same area in one cage and have the same sizing. As shown in FIG. 2A, a chicken body 40 crouching on the floor is confirmed at the bottom of the cage, indicating that it is a dead bird. On the other hand, in FIG. 2B, the legs 41 of the chicken are confirmed at the lower part of the cage, and the body of the chicken crouching on the floor is not confirmed, so that it can be seen that there are no dead birds. As described above, such an image may be used as it is as the dead bird image and the live bird image, or may be used as the learning data.

The image acquisition unit 11 acquires a plurality of scanned images (hereinafter, also referred to as estimated images) obtained by scanning the cage lanes in which the cages are continuously arranged. The scan of the cage lane is, for example, as described later, continuously capturing an image while moving spatially with respect to the cage lane in which the cage is continuous. The estimated image may be acquired by, for example, being imaged by the dead bird estimation device 1 itself, or may be acquired from an external device. In the former case, the dead bird estimation device 1 has an image pickup unit such as a camera as the image acquisition unit 11. In the latter case, the external device is, for example, an imaging device such as a camera, and the estimated image is acquired from the external device by the image acquisition unit 11 by connecting the dead bird estimation device 1 and the external device. May be good.

The connection method with the external device is not particularly limited, and it is sufficient that the estimated image can be acquired. The connection may be, for example, a wired connection or a wireless connection. The wired connection may be, for example, a cord connection or a cable connection for using a communication network. The wireless connection may be, for example, a connection using a communication network or a connection using wireless communication. The communication network is not particularly limited, and is, for example, the same as described above.

When the dead bird estimation device 1 has the imaging unit, the cage lane is continuously scanned. Therefore, the dead bird estimation device 1 is preferably a mobile device that can move in the poultry house.

In the poultry house, as described above, cage lanes in which a plurality of cages are continuous are arranged. Here, FIG. 3 shows a schematic diagram of the arrangement of cage lanes in the poultry house. Note that FIG. 3 is merely an example for explaining the present invention, and the present invention is not limited to this example. 3A and 3B are schematic views of cage lanes in the poultry house 2, FIG. 3A is a plan view of cage lanes A to F inside the poultry house 2 from above, and FIG. 3B is a plan view of the cage lanes inside the poultry house 2. 2 is a plan view of the cage lane A inside the 2 as viewed from the front direction. In FIG. 3, arrow X is in the plane direction and arrow Z is in the height direction. As shown in FIG. 3, a plurality of cage lanes A to F are arranged in the poultry house 2 along the plane direction. Each cage lane includes a cage group (row) continuous in the plane direction X and a cage group (stage) continuous in the height direction. The positions of each cage 30 are, for example, lanes (A, B, C, D, E or F), in order in the plane direction (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11). Or 12), and can be identified by the steps (1, 2, 3, 4) in the height direction. The lanes A and B, the lanes C and D, and the lanes E and F are arranged so that the back surfaces of the cages 30 face each other, for example. In FIG. 3, for convenience, the number of lanes is 6, the number of cages 30 in a row is 12, and the number of stages is 4, but this is merely an example, and the present invention is not limited thereto.

The plurality of estimated images can be scanned as follows, for example. On the front side of lane A, while moving the camera from cage A1 to A12, the cage lane A is continuously imaged, and on the front side of lane B, from cage B12 to B1. Cage lane B is continuously imaged while moving the camera, and cage lane C is continuously imaged while moving the camera from cage C1 to C12 on the front side of lane C. Cage lane D is continuously imaged while moving the camera from cage D12 to D1 on the front side of D, and the camera is continuously imaged on the front side of lane E from cage E1 to E12. The cage lane E is continuously imaged while moving the camera, and the cage lane F is continuously imaged while moving the camera from the cage F12 toward F1 on the front side of the lane F. In this way, all cage lanes can be scanned continuously.

Further, for example, a camera that captures an image in the right direction and a camera that captures an image in the left direction are used in combination to travel between cage lanes B and C and between cage lanes D and E, while traveling in the right direction and the left direction. By imaging the direction, the estimated images of the cage lanes B and C and the estimated images of the cage lanes D and E can be acquired at the same time.

Further, as shown in FIG. 3, when the cage 30 is stacked in a plurality of stages, the cage lanes A to F may be imaged for each stage, or the cage lanes A to F may be imaged at the same time. In the latter case, for example, by fixing the cameras at the heights of the cage lanes A to F at the heights of the stages 1 to 4 and taking images while proceeding, the estimated images of the stages can be acquired at the same time.

As the plurality of estimated images, for example, an image captured under the following imaging conditions is preferable. The imaging conditions are, for example, a moving speed of 2 to 8 km / hr and 5 to 35 images taken per second. The estimated image is preferably an image including, for example, the lower region of the cage for the same reason as the above-mentioned dead bird estimation information. Further, since the estimated image is finally estimated for the cage image extracted from the estimated image, as will be described later, the imaging conditions thereof are, for example, the same conditions as the image used for the dead bird estimation information. It is preferable to have.

The cage extraction unit 12 extracts cage images for each of the continuously arranged cages from the plurality of estimated images. The method of extracting the cage image for each cage from the estimated image is not particularly limited, and can be performed, for example, by using an image analysis method. Information for extracting (cutting out) the cage image from the estimated image may be stored in advance in the storage unit 10 as, for example, extraction information. The extraction information is not particularly limited, and is information related to cutting out from the estimated image. The extraction information includes, for example, imaging conditions of the estimated image (for example, camera setting conditions), position information of the region to be extracted, and the like.

When the cage image is extracted, for example, by specifying the shape of the cage in the estimated image, the information for extraction is used, for example, in a general cage image or in a poultry house using the present invention. Images of cages can be used. In this case, for example, the cage image can be extracted from the estimated image by collating the estimated image with the image of the extraction information. The method is not limited to this method, and for example, the cage feature information in the image may be used as the extraction information, and the cage image may be extracted from the estimated image based on the extraction information.

The cage image extracted from the estimated image is preferably, for example, a predetermined image of a predetermined area in the cage. In this way, by extracting an image of a predetermined area for each cage, more accurate estimation becomes possible. Further, it is preferable that the cage extraction unit 12 cuts out an image of a predetermined area in the cage, resizes the image of the cut out default area to an image of a predetermined size, and extracts the resized image as the cage image. The default size may be, for example, the same size as the image used for the above-mentioned dead bird estimation information. This also enables more accurate estimation. The predetermined region is preferably the lower region of the cage, for example, for the same reason as the dead bird estimation information. The setting condition of the default area and the setting condition of the default size may be stored in advance in the storage unit 10 as the extraction information, for example, as described above.

The dead bird estimation unit 13 estimates whether or not the bird in the cage in each cage image is a dead bird based on the dead bird estimation information for each of the extracted cage images.

When the dead bird estimation information is the dead bird estimation model which is the first example, it is possible to estimate whether or not the bird in the cage in each gauge image is a dead bird by using the dead bird estimation model. .. Further, when the dead bird estimation information is the dead bird image and the live bird image itself in the second example, the dead bird image and the live bird image are collated with each gauge image. , It is possible to estimate whether or not the bird in the cage in each gauge image is a dead bird. Further, when the dead bird estimation information is information obtained from the dead bird image and the live bird image in the second example, for example, feature point information, each gauge image is from the dead bird image. By analyzing which of the obtained feature points and the feature points obtained from the live bird image is possessed, it is possible to estimate whether or not the bird in the cage in each gauge image is a dead bird.

The output unit 14 is arbitrary and outputs the obtained estimation result. When the output unit 14 has, for example, a display unit such as a display described later, the estimation result can be output to the display unit. Further, the output unit 14 may output the estimation result to, for example, an external device. The external device is not particularly limited, and is, for example, a device having a display unit such as a display, and specific examples thereof include external terminals such as mobile phones, smartphones, tablets, and PCs. As described above, the dead bird estimation device 1 and the external device can output the estimation result from the dead bird estimation device to the external device by connecting. The connection type is not particularly limited and is the same as described above.

The method of outputting the estimation result is not particularly limited. When the dead bird estimation device 1 has the display unit, for example, the cage images of each cage are sequentially projected on the display unit, and when the cage image of the cage in which the dead bird is presumed to exist is projected, an alert is also provided. May be output. The alert may be, for example, an alert of characters, colors, blinking, etc. on the display unit, or an alert of sound.

Next, FIG. 4 illustrates a block diagram of the hardware configuration of the dead bird estimation device 1 of FIG. As shown in FIG. 4, the dead bird estimation device 1 includes, for example, a CPU (central processing unit) 101, a memory 102, an interface (I / F) 103, an input device 104, a display 105, a communication device 106, and a storage device 107. It has an image pickup device 110 and the like. Each part of the dead bird estimation device 1 is connected via, for example, I / F 103. In the hardware configuration, communication between internal circuits is connected by a bus.

The CPU 101 is responsible for overall control of the dead bird estimation device 1. In the dead bird estimation device 1, for example, the program of the present invention and other programs are executed by the CPU 101, and various information is read and written. Specifically, in the dead bird estimation device 1, for example, the CPU 101 functions as an image acquisition unit 11, a cage extraction unit 12, and a dead bird estimation unit 13. As will be described later, when the dead bird estimation model 1091 is stored as the dead bird estimation information 109 in the storage unit 10, the CPU 101 functioning as the dead bird estimation unit 13 calls the dead bird estimation model 1091 and calls the dead bird estimation model 1091. The estimation process is performed using 1091.

The I / F 103 connects, for example, between the functional units of the CPU 101, the memory 102, and the like described above. Further, the I / F 103 can also be connected to, for example, an external device. The dead bird estimation device 1 can be connected to the communication network by, for example, the communication device 106 connected to the I / F 103, and can also be connected to the external device via the communication network.

The memory 102 includes, for example, a main memory, and the main memory is also referred to as a main storage device. When the CPU 101 performs processing, for example, the memory 102 reads various programs such as the program of the present invention stored in the auxiliary storage device described later, and the CPU 101 receives data from the memory 102 and described above. Run the program. The main memory is, for example, a RAM (random access memory). The memory 102 further includes, for example, a ROM (read-only memory).

The storage device 107 is also referred to as a so-called auxiliary storage device with respect to the main memory (main storage device), for example. The storage device 107 includes, for example, a storage medium and a drive for reading and writing to the storage medium. The storage medium is not particularly limited, and may be, for example, an internal type or an external type, and may be an HD (hard disk), FD (floppy (registered trademark) disk), CD-ROM, CD-R, CD-RW, MO, etc. Examples thereof include a DVD, a flash memory, a memory card, and the like, and the drive is not particularly limited. As the storage device 107, for example, a hard disk drive (HDD) in which a storage medium and a drive are integrated, and a solid state drive (SSD) can be exemplified. The storage device 107 stores, for example, an operation program 108 such as the program of the present invention, and as described above, when the CPU 101 is executed, the memory 102 reads the operation program 108 from the storage device 107. The storage device 107 is, for example, a storage unit 10, in which the dead bird estimation model 1091 is stored as the dead bird estimation information 109, and information acquired in each unit, information used for processing in each unit, and the like are stored.

The dead bird estimation device 1 may further include an image pickup device 110. The image pickup device 110 is, for example, a camera. When the dead bird estimation device 1 includes the image pickup device 110, for example, the dead bird estimation device 1 is moved in the poultry house, and the cage lane is imaged by the image pickup device 110 to move in the poultry house and die in real time. It is also possible to estimate the existence of birds.

The dead bird estimation device 1 further optionally has, for example, an input device 104 and a display 105. The input device 104 is, for example, a touch panel, a keyboard, a mouse, or the like. The display 105 is, for example, an LED display, a liquid crystal display, or the like.

Further, when the dead bird estimation device 1 further includes the learning unit that generates the dead bird estimation model 1091, as described above, the block diagram of the hardware configuration of FIG. 5 can be exemplified.

As shown in FIG. 5, in the dead bird estimation device 1, for example, the storage device 107 further stores the learning data 111 including the live bird image information 1111 and the dead bird image information 1112. The CPU 101 further functions as a learning unit 14, takes out the learning data 111 of the storage device 107, generates a dead bird estimation model 1091 by learning, and the storage device 107 stores it as the dead bird estimation information 109. ..

Next, the dead bird estimation method of the present embodiment will be described with reference to the flowchart of FIG. The information providing method of the present invention is not limited to these examples. The dead bird estimation method of the present embodiment can be carried out using, for example, the dead bird estimation device 1 shown in FIG. 1, but the dead bird estimation method of the present embodiment is not limited to the use of the dead bird estimation device 1.

The dead bird estimation model is stored in advance in the storage unit 10 of the dead bird estimation device 1 as the dead bird estimation information. Further, as described above, the storage unit 10 also stores, for example, the extraction information.

First, a plurality of estimated images obtained by continuously scanning the cage lanes in which the cages are continuously arranged are acquired (S100). This step can be executed, for example, by the image acquisition unit 11 of the dead bird estimation device 1.

Next, the cage images for each of the continuously arranged cages are extracted from the plurality of estimated images. This step can be performed, for example, by the cage extraction unit 12 of the dead bird estimation device 1. The cage image can be extracted as a resized cage image, for example, by cutting out an image of the defined region from the estimated image (S101) and resizing the image of the cut out default region (S102).

Subsequently, with respect to the extracted cage image, it is determined whether or not the bird in the cage in each cage image is a dead bird based on the dead bird estimation information (S103). Specifically, by using the dead bird determination model generated by learning, it is possible to estimate whether or not a dead bird exists for each cage image. When it is determined that the bird is a dead bird (YES), it is presumed that a dead bird exists in the cage corresponding to the cage image, and when it is determined that the bird is not a dead bird (NO), the cage corresponding to the cage image is used. It is estimated that there are no dead birds. Then, these results are output (S106), and the process ends. The step (S101) to the step (S106) are executed for each of the estimated images.

(Modification 1)
Although the flowchart of FIG. 6 shows an example using the dead bird estimation model, the present invention is not limited thereto. When the dead bird image and the live bird image itself are used as the dead bird estimation information in the step (S103), for example, the dead bird image and the live bird image are collated with each gauge image. By doing so, it is possible to estimate whether or not the bird in the cage in each gauge image is a dead bird. When information obtained from the dead bird image and the live bird image, for example, information on feature points is used as the dead bird estimation information, for example, each gauge image can be obtained from the dead bird image. By analyzing which of the feature points and the feature points obtained from the live bird image is possessed, it is possible to estimate whether or not the bird in the cage in each gauge image is a dead bird.

In this case, the storage unit 10 of the storage device 107 stores, for example, the dead bird image information and the live bird image information as the dead bird estimation information.

[Embodiment 2]
The present embodiment is another example of the dead bird estimation device and the dead bird estimation method of the present invention, and further is an embodiment for specifying a cage corresponding to the cage image. FIG. 7 shows a block diagram of an example of the dead bird estimation device of the present embodiment.

The dead bird estimation device 3 of the present embodiment is the same as that of the first embodiment except that the storage unit 10 further includes the cage specific information storage unit 102 and the cage specific information storage unit 15. The cage specific information storage unit 102 stores specific information that identifies the cage in the image, and the cage identification unit 15 corresponds to each estimated image or each extracted cage image in the poultry house by the specific information. Identify the cage.

The dead bird estimation method of the second embodiment is the same as that of the first embodiment except that the cage specifying step is further included. In the cage specifying step, the corresponding cage in the poultry house is specified by the specific information for each estimated image or each extracted cage image.

The specific information that identifies the cage in the image is not particularly limited. The first example of the specific information is, for example, the identification information of each cage in the poultry house. When each cage in the poultry house has a different ID code in front of the cage, for example, the ID code and the position information of the cage associated with the ID code (for example, "A1 / 1" shown in FIG. 3B). "Etc.)" may be stored as identification information. In this case, an ID code is extracted from the estimated image or the cage image, the extracted ID code is collated with the ID code stored in the storage unit, and the cage of the matching ID code is the estimated image. Alternatively, it can be identified as a cage corresponding to the cage image. Further, since the position information is associated with the ID code, for example, the position information of the cage in the actual poultry house can also be obtained from the ID code.

A second example of the specific information is, for example, the address information of each cage. As shown in FIG. 3A, the poultry house 2 has a plurality of lanes (for example, A to F) in which a plurality of cages are continuous. Then, when a plurality of images are continuously acquired along the lane and cages are extracted from the plurality of images in that order, it is a continuous cage in the lane. Specifically, for example, if the cage is imaged in the direction of the arrow X for the lane A and the cages are extracted in the order of the acquired plurality of images, the extracted first cage image is the cage A1. The image of the above and the extracted second cage image are the images of the cage A2. In this case, the images to be continuously acquired are, for example, images including the left and right sides of the cage, so that each cage can be extracted from the left and right divisions of the cage and the like. The same applies when the cages are stacked in the upward direction. For example, when four cameras are arranged according to the cages having one to four stages, the first, second, and third from the bottom. By associating the ID of the first and fourth cameras with the image, the cage at which stage is also determined. In this case, by making the image including the upper and lower parts of the cage, each cage can be extracted from the upper and lower divisions of the cage and the like. Therefore, images for each lane are acquired in advance, cage images are extracted in order from the images, and cage addresses, for example, A1 and A1 shown in FIGS. 3A and 3B, are obtained according to the order. Address information such as / 1 can be set. Then, in the actual dead bird estimation, similarly, the estimated image for each lane is acquired, the cage image is sequentially extracted from the estimated image, and the address set in advance as the specific information according to the order. Information can be given as address information for each cage image. This makes it possible to identify which cage in the poultry house corresponds to each cage image extracted from the estimated image from the given address information.

As the second example, the specific information includes, for example, a cage specific model. The cage-specific model is, for example, a model that represents a relationship (algorithm) between an image and a cage in the image. The cage specific model is generated by learning, for example, using a cage image as learning data. The learning includes so-called machine learning, and the above-mentioned system can be used. In this case, since the cage image can be extracted from the estimated image by using the cage specific model, the cage address can be given to the continuous cage images in that order as described above.

For setting the address information for each cage, for example, the information of the poultry house identification information (name, number, etc.), the lane identification information (lane number, the number of stages, etc.), and the camera identification information (ID, etc.) are further added. It is preferable to associate them.

According to the second embodiment, the presence or absence of a dead bird in each cage image can be estimated, and which cage in the poultry house the cage image to be estimated corresponds to can be specified. Therefore, the worker can easily check the inside of the target cage by going to the cage where the dead bird is presumed to be present, for example.

[Embodiment 3]
The third embodiment is another example of the dead bird estimation device and the dead bird estimation method of the present invention, and is a mobile device provided with a camera as an image pickup unit as described above.

An outline example of the dead bird estimation device according to the third embodiment is shown in FIG. FIG. 8A is a side view of the dead bird estimation device 5, and FIG. 8B is a front view of the dead bird estimation device 5.

In FIG. 8A, the arrow indicates the traveling direction of the dead bird estimation device 5. As shown in FIG. 8, the dead bird estimation device 5 includes a wheel portion 51, a base portion 52, a grip portion 53, a PC (personal computer) 54, a battery 55, a USB hub 56, a support portion 57, a patrol lamp 58, and cameras 111 to. It is equipped with 114. A wheel portion 51 is installed below the base portion 52, a grip portion 53 for workers is arranged on the front side, and a support portion 57 is arranged on the traveling direction side on the base portion 52. Four sets (111 to 114) of a pair of cameras are fixed to the support portion 57 according to the height of each stage of the cage. The pair of cameras 111 to 114 are fixed to the support portion 57 so as to face each other in opposite directions so that the right direction (R) and the left direction (L) can be simultaneously imaged. A patrol lamp is arranged at the bottom of each pair of cameras 111 to 114. A PC 54 and a battery 55 are mounted on the base portion 52, and a USB hub 56 is optionally arranged in addition to the PC 54 and the battery 55.

According to the dead bird estimation device 5, for example, as described in the first embodiment, the right lane and the left lane are simultaneously passed while passing between the lane B and the lane C shown in FIG. You can take an image. Further, since the cameras 111 to 114 are fixed to the four stages, it is possible to simultaneously image the first to fourth stages of the cage lane. Then, the estimated images captured by the cameras 111 to 114 are image-processed by the mounted PC 54 to estimate dead birds, and if it is estimated that there are dead birds, the patrol lamp near the corresponding cage is used in real time. By turning on, it is possible to inform the worker of the possibility of the existence of a dead bird. In this way, the worker can know the possibility of the existence of the dead bird in real time while pushing the dead bird estimation device 5, so that, for example, the dead bird is confirmed and removed from the target cage on the spot. can do. Further, the dead bird estimation device 5 is, for example, moved by an operator or automatically run, and the obtained estimation result is stored in a storage unit. Later, after confirming the estimation result, the worker moves the poultry house. It is also possible to go to and identify and remove dead birds only in the target cage.

As shown in the first embodiment, the estimated image preferably includes an image of the lower region of the cage, that is, the foot of the chicken. Therefore, in the dead bird estimation device 5, it is preferable that the cameras 111 to 114 are fixed at an angle at which only the legs of the chicken are photographed with respect to the cage to be imaged. As mentioned above, the cage generally has its floor tilted toward the front for egg recovery. Therefore, for example, as shown in the schematic view of FIG. 9, it is preferable to fix the camera 111 to the cage 30 having the inclined floor 31 so that the legs of the chicken 42 can be imaged.

[Embodiment 4]
The program according to the fourth embodiment of the present invention is a program capable of executing the dead bird estimation method of the present invention on a computer. Alternatively, the program of this embodiment may be recorded on a computer-readable recording medium, for example. The recording medium is not particularly limited, and examples thereof include a storage medium as described above.

Although the invention of the present application has been described above with reference to the embodiments, the invention of the present application is not limited to the above-described embodiment. Various changes that can be understood by those skilled in the art can be made within the scope of the present invention in terms of the configuration and details of the present invention.

<Additional Notes>
Some or all of the above embodiments and examples may be described as, but not limited to, the following appendixes.
(Appendix 1)
Includes image acquisition section, cage extraction section, and dead bird estimation section
The image acquisition unit
Acquire multiple scan images of scanned cage lanes in which cages are continuously arranged,
The cage extraction unit
Cage images for each of the continuously arranged cages are extracted from the plurality of scanned images, and the dead bird estimation unit is used.
For each of the extracted cage images, it is estimated whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation information for estimating whether or not the bird in the cage in the image is a dead bird.
A dead bird estimation device characterized by this.
(Appendix 2)
The dead bird estimation information is a dead bird estimation model that estimates whether or not the bird in the cage in the image is a dead bird.
The dead bird estimation model is a model generated by learning from a dead bird image of a dead bird in a cage and a live bird image of a live bird in the cage.
The dead bird estimation unit according to Appendix 1, wherein the dead bird estimation unit estimates whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation model.
(Appendix 3) The dead bird estimation information is the dead bird image information indicating the state of the dead bird in the cage and the live bird image information indicating the state of the live bird in the cage.
The dead bird estimation device according to Appendix 1, wherein the dead bird estimation unit estimates whether or not the bird in the cage in each cage image is a dead bird based on the dead bird image information and the live bird image.
(Appendix 4)
The dead bird estimation device according to Appendix 1, further comprising a cage identification unit that identifies a corresponding cage in a poultry house by means of specific information that identifies a cage in the image for each scanned image or each extracted cage image.
(Appendix 5)
The specific information is a cage specific model that identifies the cage in the image.
The cage specific model is a model generated by learning from the images of each cage.
The dead bird estimation device according to Appendix 4, wherein the cage specifying portion identifies the cage of each scanned image or each cage image by the cage specifying model.
(Appendix 6)
The dead bird estimation device according to any one of Supplementary note 1 to 5, wherein the cage extraction unit extracts an image of a predetermined area in the cage as the cage image.
(Appendix 7)
The cage extraction unit extracts an image of a default area in the cage, resizes the extracted image of the default area to an image of a default size, and uses the resized image as the cage image. The dead bird estimation device described in.
(Appendix 8)
The dead bird estimation device according to Appendix 6 or 7, wherein the image of the predetermined area is an image of the area where the legs of the bird in the cage are located.
(Appendix 9)
The dead bird estimation device according to any one of Supplementary note 1 to 8, wherein the image acquisition unit is an image pickup unit.
(Appendix 10)
The dead bird estimation device according to any one of Supplementary note 1 to 9, further comprising an output unit that outputs the estimation result.
(Appendix 11)
Including image acquisition process, cage extraction process, and dead bird estimation process
The image acquisition step is
Acquire multiple scan images of scanned cage lanes in which cages are continuously arranged,
The cage extraction step is
From the plurality of scanned images, cage images for each of the continuously arranged cages are extracted.
The dead bird estimation process is
For each of the extracted cage images, it is characterized in that it is estimated whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation information for estimating whether or not the bird in the cage in the image is a dead bird. Dead bird estimation method.
(Appendix 12)
The dead bird estimation information is a dead bird estimation model that estimates whether or not the bird in the cage in the image is a dead bird.
The dead bird estimation model is a model generated by learning from a dead bird image of a dead bird in a cage and a live bird image of a live bird in the cage.
The dead bird estimation method according to Appendix 11, wherein the dead bird estimation step estimates whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation model.
(Appendix 13)
The dead bird estimation information is the dead bird image information showing the state of the dead bird in the cage and the live bird image information showing the state of the live bird in the cage.
The dead bird estimation method according to Appendix 11, wherein the dead bird estimation step estimates whether or not the bird in the cage in each cage image is a dead bird based on the dead bird image information and the live bird image.
(Appendix 14)
Further, the cage identification step of identifying the corresponding cage in the poultry house by the specific information for identifying the cage in the image for each scanned image or the extracted cage image is included.
The dead bird estimation method according to Appendix 11.
(Appendix 15)
The specific information is a cage specific model that identifies the cage in the image.
The cage specific model is a model generated by learning from the images of each cage.
The dead bird estimation method according to Appendix 13, wherein the cage identification step identifies the cage of each scanned image or each cage image by the cage identification model.
(Appendix 16)
The dead bird estimation method according to any one of Supplementary note 11 to 15, wherein the cage extraction step extracts an image of a predetermined area in the cage as the cage image.
(Appendix 17)
The cage extraction step extracts an image of a predetermined area in the cage, resizes the image of the extracted default area to an image of a predetermined size, and uses the resized image as the cage image. The dead bird estimation method described in.
(Appendix 18)
The dead bird estimation method according to Appendix 16 or 17, wherein the image of the predetermined area is an image of the area where the leg of the bird is located in the cage.
(Appendix 19)
The dead bird estimation method according to any one of Supplementary note 11 to 18, wherein the image acquisition step is an imaging step.
(Appendix 20)
The dead bird estimation method according to any one of Supplementary note 11 to 19, further comprising an output step of outputting the estimation result.
(Appendix 21)
A program comprising causing a computer to execute the dead bird estimation method according to any one of Supplementary note 11 to 20.
(Appendix 22)
A computer-readable recording medium on which the program described in Appendix 21 is recorded.

According to the present invention, it is possible to easily estimate dead birds in a cage in real time.

1 Dead bird estimation device 10 Storage unit 101 Dead bird estimation information storage unit 11 Image acquisition unit 12 Cage extraction unit 13 Dead bird estimation unit 14 Output unit

Claims (10)

Includes image acquisition section, cage extraction section, and dead bird estimation section
The image acquisition unit
Acquire multiple scan images of scanned cage lanes in which cages are continuously arranged,
The cage extraction unit
Cage images for each of the continuously arranged cages are extracted from the plurality of scanned images, and the dead bird estimation unit is used.
For each of the extracted cage images, it is estimated whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation information for estimating whether or not the bird in the cage in the image is a dead bird.
A dead bird estimation device characterized by this. The dead bird estimation information is a dead bird estimation model that estimates whether or not the bird in the cage in the image is a dead bird.
The dead bird estimation model is a model generated by learning from a dead bird image of a dead bird in a cage and a live bird image of a live bird in the cage.
The dead bird estimation device according to claim 1, wherein the dead bird estimation unit estimates whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation model. The dead bird estimation information is the dead bird image information showing the state of the dead bird in the cage and the live bird image information showing the state of the live bird in the cage.
The dead bird estimation device according to claim 1, wherein the dead bird estimation unit estimates whether or not the bird in the cage in each cage image is a dead bird based on the dead bird image information and the live bird image information. The dead bird estimation device according to claim 1, further comprising a cage identification unit that identifies a corresponding cage in a poultry house by means of specific information that identifies a cage in the image for each scanned image or each extracted cage image. .. The specific information is a cage specific model that identifies the cage in the image.
The cage specific model is a model generated by learning from the images of each cage.
The dead bird estimation device according to claim 4, wherein the cage identification unit identifies the cage of each of the scanned images or the cage images by the cage identification model. The dead bird estimation device according to any one of claims 1 to 5, wherein the cage extraction unit extracts an image of a predetermined area in the cage as the cage image. Any of claims 1 to 6, wherein the cage extraction unit extracts an image of a default area in the cage, resizes the extracted image of the default area to an image of a default size, and uses the resized image as the cage image. The dead bird estimation device described in item 1. The dead bird estimation device according to claim 6 or 7, wherein the image of the predetermined area is an image of the area where the foot of the bird in the cage is located. Including image acquisition process, cage extraction process, and dead bird estimation process
The image acquisition step is
Acquire multiple scan images of scanned cage lanes in which cages are continuously arranged,
The cage extraction step is
From the plurality of scanned images, cage images for each of the continuously arranged cages are extracted.
The dead bird estimation process is
For each of the extracted cage images, it is estimated whether or not the bird in the cage in each cage image is a dead bird by the dead bird estimation information for estimating whether or not the bird in the cage in the image is a dead bird.
A dead bird estimation method characterized by this. A program comprising causing a computer to execute the dead bird estimation method according to claim 9.

Images