JP2019205425A - Dead chicken sensing system, chicken life or death determination program, and chicken life or death determination device - Google Patents

Abstract

To provide a dead chicken sensing system capable of detecting dead chickens with high accuracy.SOLUTION: A dead chicken sensing system 1 includes a camera 13 for photographing chickens in a cage, and a chicken life or death determination device 30 for determining the life or death of the chickens based on whether or not the dead chickens are contained in a monitoring image acquired from the camera 13. The chicken life or death determination device 30 constitutes a learning instrument where machine learning is performed previously by using the learning data containing a life image being the image obtained by photographing the chickens in the cage and not containing the dead chickens and a death image containing the dead chickens. The learning instrument is used in which the machine learning is performed by using the learning data containing the life image and the death image, and thereby the dead chickens can be detected with high accuracy.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a dead chicken detection system, a chicken life / death determination program, and a chicken life / death determination apparatus. More specifically, the present invention relates to a system for detecting dead chickens in a poultry farm, and a program and an apparatus for determining whether chickens are alive or dead.

  In the poultry farm for egg laying chickens, several chickens are raised in cages. When a chicken in a cage dies, eggs that have been laid by other chickens are caught by the dead chicken (hereinafter referred to as “dead chickens”) and are not ejected from the cage and may rot. Also, other chickens may become sick due to rotting eggs and rotting dead chickens. This will cause economic damage to the poultry farmer. In order to avoid such a situation, when a chicken dies, it is necessary to quickly remove the chicken from the cage.

  Traditionally, the discovery of dead chickens has been done manually. In other words, poultry farm staff regularly monitored the poultry house and removed dead chickens when they were found. However, there is a problem that such human work is a heavy burden on the staff.

  With respect to this problem, Patent Document 1 discloses an apparatus that automatically detects dead animals. This apparatus uses a two-dimensional laser scanner to periodically map the outlines of animals in the breeding ground, and compares successive mappings to detect a stationary animal with no change in outline as a dead animal. If this device is used, dead chickens can be automatically detected, reducing the burden on the staff.

Special table 2017-505639 gazette

  The device of Patent Document 1 detects a stationary animal as a dead animal on the assumption that a healthy animal is always moving. However, in reality, even a live chicken may be stationary. For example, a live chicken may be sitting in the same place. In particular, older chickens tend to sit longer. Also, dead chickens are not necessarily stationary. Chickens have the habit of kicking away foreign objects around them. Due to this behavior, dead chickens may be moved by other chickens. Therefore, the apparatus of Patent Document 1 may erroneously detect a live chicken as a dead chicken, and may not detect a dead chicken.

An object of this invention is to provide the dead chicken detection system which can detect a dead chicken accurately in view of the said situation.
Another object of the present invention is to provide a chicken life / death determination program and a chicken life / death determination apparatus that can accurately determine the life / death of a chicken.

A dead chicken detection system according to a first aspect of the present invention is a life / death determination for chickens that determines whether the chickens are alive or not based on whether or not a dead chicken is included in a monitoring image acquired from the camera that captures the chickens in the cage. And a device.
The dead chicken detection system according to a second aspect of the present invention is the first aspect of the invention, wherein the chicken life / death determining device includes an image obtained by photographing a chicken in a cage and does not include a dead chicken, and a dead chicken. A learning device in which machine learning is performed in advance is configured using learning data including a death image.
The dead chicken detection system according to a third aspect of the present invention is the second aspect of the invention, wherein the survival image includes a standing image in which all chickens in the image are standing, and a part of the chickens in the image are standing and the remaining chickens are A sitting image is included.
According to a fourth aspect of the present invention, in the second or third aspect, the dead image includes a fallen image in which some chickens in the image have fallen, a defect in which some chickens in the image have a defect. One or more types of images and discolored images in which some of the chickens in the image are discolored are included.
According to a fifth aspect of the present invention, in the second or third aspect, the death image includes the type of dead chicken, the position of the dead chicken in the cage, the angle of the dead chicken, the direction of the dead chicken, the cage Among the above-described configurations, a plurality of patterns of images obtained by shooting one or more of the images are included.
In any one of the second to fifth inventions, the chicken life / death determining device determines that the dead chicken detection system of the sixth invention is dead when the probability that a dead chicken is included in the monitoring image is equal to or greater than a death threshold. When the probability that a dead chicken is included in the monitoring image is equal to or less than a survival threshold, it is determined to be survival, and the probability that a dead chicken is included in the monitoring image exceeds the survival threshold and is less than the death threshold. In addition, it is characterized in that it is determined that life and death are unknown.
According to a seventh aspect of the present invention, there is provided a dead chicken detection system according to the sixth aspect, wherein the chicken life / death determining apparatus determines that all are unidentified based on the plurality of monitoring images obtained by repeatedly photographing the cage at the same position. In this case, it is characterized by re-determining that it is dead.
The dead chicken detection system according to an eighth aspect of the present invention includes the thermography camera for photographing the chicken in the cage according to the sixth aspect, wherein the chicken life / death determining apparatus determines that the life / death is unknown based on the monitoring image. A temperature of a chicken region whose life and death is unknown among temperature images acquired from a thermographic camera is acquired, and when the temperature is equal to or lower than a temperature threshold, it is determined again as dead.
According to a ninth aspect of the present invention, in the first aspect of the present invention, the chicken life / death determining apparatus is configured such that all the chickens in the monitoring image are standing or a part of the chickens in the monitoring image are standing. When the remaining chickens are sitting, it is determined that dead chickens are not included in the monitoring image.
According to a tenth aspect of the present invention, in the first or ninth aspect, the chicken alive / death determining device is configured such that a part of chickens in the monitoring image has fallen, and a part of the chickens in the monitoring image is deficient. Or when a part of chickens in the monitoring image is discolored, it is determined that a dead chicken is included in the monitoring image.
The dead chicken detection system according to an eleventh aspect of the invention is characterized in that, in any one of the first to tenth aspects of the invention, the chicken life / death determining apparatus notifies dead chicken information when it is determined to be dead.
The dead chicken detection system according to a twelfth aspect of the invention is characterized in that, in the eleventh aspect of the invention, the dead chicken information includes the monitoring image that is the basis for the death determination.
The dead chicken detection system of the thirteenth aspect of the invention is characterized in that, in the eleventh or twelfth aspect of the invention, the dead chicken information includes position information of the cage where the monitoring image that is the basis of death determination is taken. To do.
A dead chicken detection system according to a fourteenth aspect of the invention is characterized in that, in any one of the first to thirteenth aspects, illumination having a color temperature of 4,000 to 7,000 K for illuminating the photographing range of the camera is provided.
The fowl detection system of the fifteenth aspect of the present invention is characterized in that, in the fourteenth aspect, the illumination is always on.
A dead chicken detection system according to a sixteenth aspect of the present invention is characterized in that, in any one of the first to fifteenth aspects, the camera is mounted and a traveling carriage that travels along the cage is provided.
A chicken life / death determination program according to a seventeenth aspect of the invention is a chicken life / death determination program for causing a computer to function based on a monitoring image obtained by photographing a chicken in a cage to determine the life / death of the chicken. A learning machine that has been pre-machine-learned is constructed using learning data that includes images of live images that do not include dead chickens and images that are obtained by photographing chickens in them, and death images that contain dead chickens. When the monitoring image is input to the learning device, the computer is caused to function so as to output a determination result as to whether or not a dead chicken is included in the monitoring image.
According to an eighteenth aspect of the present invention, in the seventeenth aspect, the survival image includes a standing image in which all the chickens in the image are standing, and a part of the chickens in the image and the remaining chickens. A sitting image is included.
According to a nineteenth aspect of the present invention, in the seventeenth or eighteenth aspect, the death image includes a fallen image in which some chickens in the image have fallen, and a defect in which some chickens in the image have a defect. One or more types of images and discolored images in which some of the chickens in the image are discolored are included.
According to a twentieth aspect of the present invention, in the seventeenth or eighteenth aspect, the death image includes the type of dead chicken, the position of the dead chicken in the cage, the angle of the dead chicken, the direction of the dead chicken, the cage Among the above-described configurations, a plurality of patterns of images obtained by shooting one or more of the images are included.
In any one of the seventeenth to twentieth inventions, the chicken life / death determination program according to the twenty-first aspect of the invention determines that the death is included in the monitoring image when the probability that the monitoring image includes a dead chicken is equal to or greater than a death threshold. When the probability that a chicken is included is less than or equal to the survival threshold, it is determined to be live, and when the probability that a dead chicken is included in the monitoring image exceeds the survival threshold and is less than the death threshold, it is determined that life or death is unknown. As described above, the computer is made to function.
In the twenty-first invention, the chicken life / death determination program according to the twenty-second invention is based on the plurality of monitoring images obtained by repeatedly photographing the cage at the same position. It is characterized by making a computer function so as to make a determination.
In the twenty-third invention, the life / death determination program for chickens in the twenty-first invention, when it is determined that the life / death is unknown based on the monitoring image, the life / death is unknown among the temperature images obtained by photographing the chicken in the cage with a thermography camera. The temperature of the chicken area is acquired, and the computer is caused to function so as to re-determine the death when the temperature is equal to or lower than the temperature threshold.
A chicken life / death determination program according to a twenty-fourth aspect of the invention is a chicken life / death determination program for causing a computer to function based on a monitoring image obtained by photographing a chicken in a cage to determine the life / death of the chicken. When all the chickens in the monitoring image are standing, or when a part of the chickens in the monitoring image are standing and the remaining chickens are sitting, the computer is made to function so as to be determined to be alive. And
A chicken life / death determination program according to a twenty-fifth aspect of the invention is a chicken life / death determination program for causing a computer to function based on a monitoring image obtained by photographing a chicken in a cage to determine the life / death of the chicken. If some chickens in the monitoring image are fallen, some chickens in the monitoring image are defective, or some chickens in the monitoring image are discolored, it is determined that the chicken is dead As described above, the computer is made to function.
According to a twenty-sixth aspect of the present invention, there is provided a chicken life / death determining apparatus comprising a computer in which the chicken life / death determination program according to any of the seventeenth to twenty-fifth aspects is installed.

According to the first aspect of the present invention, since the live or dead chicken is determined based on whether or not the dead chicken is included in the monitoring image, the dead chicken can be detected with high accuracy.
According to the second invention, it is possible to accurately determine whether or not a dead chicken is included in the monitoring image by using the learning device in which machine learning is performed using the learning data including the survival image and the death image.
According to the third invention, machine learning is performed using learning data including a sitting image in addition to a standing image as a living image, so that the frequency of misjudging a chicken in which a learning device is sitting as a dead chicken is reduced. it can.
According to the fourth invention, since the machine learning is performed using the learning data including the fallen image having the characteristics of the dead chicken, the missing image, and the discoloration image as the death image, the learning device lives the dead chicken. The frequency of misjudging as a chicken can be reduced.
According to the fifth aspect, since machine learning is performed using death images of various patterns, it is possible to reduce the misjudgment rate of the learning device.
According to the sixth aspect of the present invention, the life and death of the chicken is determined in three stages: death, survival and life / death unknown. Therefore, the death / survival is not forcibly determined based on the monitoring image in which it is difficult to determine the life / death. The frequency of detection can be reduced.
According to the seventh invention, when it is determined that the life and death are continuously unknown for the monitoring image obtained by photographing the cage at the same position, the detection omission of the dead chicken is reduced by re-determination as dead. it can.
According to the eighth invention, when it is determined that the life / death is unknown based on the monitoring image, the life / death determination is performed again based on the temperature image, so that the number of cases processed as life / death unknown can be reduced.
According to the ninth aspect, in addition to the case where all the chickens in the monitoring image are standing, even when some chickens are sitting, it is determined that the dead chicken is not included in the monitoring image. This can reduce the frequency of misjudging existing chickens as dead chickens.
According to the tenth invention, when the chicken in the monitoring image has the fall, defect, or discoloration characteristic of the dead chicken, it is determined that the dead chicken is included in the monitoring image. The frequency of misjudging can be reduced.
According to the eleventh aspect, by notifying dead chicken information, it is possible to prompt the staff to remove the dead chicken.
According to the twelfth aspect of the present invention, the final determination of whether a chicken is alive or dead can be made by notifying the monitoring image that is the basis for the death determination.
According to the thirteenth invention, since the cage position information is notified, the staff can grasp the position of the dead chicken, and the removal work of the dead chicken becomes easy.
According to the fourteenth aspect, since the color of the illumination light is close to white, the stress applied to the chicken is small, and the chicken color is easily reflected in the monitoring image.
According to the fifteenth aspect of the present invention, since the lighting is always on without flashing, the stress given to the chicken can be reduced.
According to the sixteenth aspect, since the traveling carriage on which the camera is mounted travels along the cage, the entire cage can be monitored with a small number of cameras.
According to the seventeenth aspect, since the learning device in which the machine learning is performed using the learning data including the survival image and the death image is configured, it is possible to accurately determine whether the monitoring image includes a dead chicken. As a result, it is possible to accurately determine whether a chicken is alive or dead.
According to the eighteenth aspect, since machine learning is performed using learning data including a sitting image in addition to a standing image as a living image, the frequency of misjudging a chicken on which a learning device is sitting as a dead chicken is reduced. it can.
According to the nineteenth aspect of the invention, machine learning is performed using learning data including a fallen image having a characteristic of a dead chicken, a missing image, and a discolored image as a death image. The frequency of misjudging as a chicken can be reduced.
According to the twentieth aspect, since machine learning is performed using death images of various patterns, the erroneous determination rate of the learning device can be reduced.
According to the twenty-first invention, since the life and death of chickens are determined in three stages: death, survival, and unknown life and death, there is no need to forcibly determine death and survival based on a monitoring image that is difficult to determine whether or not life and death. The frequency of detection can be reduced.
According to the twenty-second invention, when it is determined that the life and death are continuously unknown with respect to the monitoring image obtained by photographing the cage at the same position, the detection omission of the dead chicken is reduced by re-determination as dead. it can.
According to the twenty-third aspect, when it is determined that the life / death is unknown based on the monitoring image, the life / death determination is performed again based on the temperature image, so the number of cases processed as life / death unknown can be reduced.
According to the twenty-fourth invention, since all the chickens in the monitoring image are standing, even when some chickens are sitting, it is determined that they are alive. Can be reduced.
According to the twenty-fifth invention, when the chicken in the monitoring image has fallen, missing, or discolored, which is a characteristic of the dead chicken, it is judged as dead, so the frequency of misjudging a dead chicken as a live chicken is reduced. it can.
According to the twenty-sixth aspect of the present invention, it is possible to accurately determine whether or not a dead chicken is included in a monitoring image by using a learning device in which machine learning is performed using learning data including a survival image and a death image. As a result, it is possible to accurately determine whether a chicken is alive or dead.

It is a whole block diagram of the dead chicken detection system which concerns on 1st Embodiment. It is a side view of the patrol robot in a 1st embodiment. It is a top view of the poultry house which shows the route of the patrol robot in a 1st embodiment. It is a block diagram of the chicken life / death determination apparatus in 1st Embodiment. It is a flowchart which shows the process by the dead chicken detection system in 1st Embodiment. It is a flowchart which shows the detail of the life-and-death determination process in 1st Embodiment. It is an example of a standing image. It is an example of a sitting image. It is an example of a fallen image. FIG. (A) is an example of a missing image when the neck of a chicken is missing. FIG. (B) is an example of a missing image when chicken wings are peeled. FIG. (A) is an example of a discolored image when a part of a chicken body is discolored. FIG. (B) is an example of a discolored image when blood is attached to a chicken. It is explanatory drawing of the position in the cage of a dead chicken. FIG. (A) is an explanatory diagram in which the angle of the dead chicken is 0 °. FIG. (B) is an explanatory diagram in which the angle of the dead chicken is 180 °. FIG. (A) is an explanatory diagram in which the direction of the dead chicken is facing left. FIG. (B) is an explanatory diagram in which the direction of the dead chicken is rightward. It is a flowchart which shows the detail of the life-and-death determination process in 3rd Embodiment. It is a whole block diagram of the dead chicken detection system which concerns on 4th Embodiment. It is a block diagram of the chicken life / death determination apparatus in 4th Embodiment. It is a flowchart which shows the detail of the life-and-death determination process in 4th Embodiment. FIG. (A) is an example of a monitoring image. FIG. (B) is an example of a temperature image corresponding to the monitoring image of FIG.

Next, an embodiment of the present invention will be described with reference to the drawings.
[First Embodiment]
The dead chicken detection system 1 according to the first embodiment of the present invention is a system for automatically detecting dead chickens in a poultry farm for egg collection chickens.

(Overall configuration of dead chicken detection system)
First, the whole structure of the dead chicken detection system 1 of this embodiment is demonstrated.
As shown in FIG. 1, the dead chicken detection system 1 includes a patrol robot 10. The traveling robot 10 is a robot for traveling around the poultry house and photographing the chickens in the cage. The number of patrol robots 10 is not particularly limited, and may be one or more.

  The traveling robot 10 has a traveling carriage 12 and a camera 13. The number of cameras 13 mounted on one traveling robot 10 is not particularly limited, and may be one or plural. The traveling carriage 12 and the camera 13 are connected to an in-flight network 11 built inside the traveling robot 10. The in-flight network 11 is not particularly limited, but is, for example, a LAN (Local Area Network). Therefore, a network camera is preferably used as the camera 13.

  An in-house network 21 is built in the poultry farm. The on-site network 21 is not particularly limited, but is, for example, a LAN (Local Area Network). A patrol robot control device 22 and a chicken life / death determination device 30 are connected to the in-site network 21.

  In addition, a wireless master device 23 is connected to the on-site network 21, and a wireless slave device 14 is connected to the on-board network 11. The wireless master device 23 and the wireless slave device 14 can perform wireless communication in both directions, and wirelessly connect the on-site network 21 and the on-board network 11. Therefore, the patrol robot control device 22 and the chicken life / death determination device 30 can communicate with the traveling carriage 12 and the camera 13 of the patrol robot 10 in both directions.

  The traveling robot control device 22 has a function of controlling the traveling of the traveling robot 10 and the like. The traveling robot control device 22 may be configured by a PLC (Programmable Logic Controller), or may be configured by a computer configured by a CPU, a memory, and the like.

  The camera 13 of the traveling robot 10 photographs the chicken in the cage. Hereinafter, an image obtained by the camera 13 photographing a chicken in the cage is referred to as a “monitoring image”. The monitoring image is transmitted to the chicken life / death determining apparatus 30 via the networks 11 and 21. The chicken life / death determining apparatus 30 determines whether or not a dead chicken is included in the monitoring image acquired from the camera 13, and determines the life / death of the chicken based on the result.

  The on-site network 21 is connected to the external network 41 via the communication control device 24. The external network 41 is not particularly limited, but is the Internet, for example. The external network 41 is connected to a terminal 40 possessed by a chicken farm staff. The terminal 40 may be a mobile terminal such as a smartphone, a tablet terminal, a mobile phone, or a general-purpose computer. The number of terminals 40 is not particularly limited, and may be one or more.

  When detecting a dead chicken, the chicken life / death determining apparatus 30 transmits information on the dead chicken (hereinafter referred to as “dead chicken information”) to the terminal 40. This notifies the chicken farm staff of dead chicken information. The transmission of dead chicken information is performed, for example, by transmitting an e-mail. Personnel who have received dead chicken information will be responsible for removing dead chickens from the cage.

(Traveling robot)
Next, details of the traveling robot 10 will be described.
As shown in FIG. 2, a plurality of cages 51 are installed in the poultry house. A plurality of chickens are bred in each cage 51. In general, the cages 51 are connected in the horizontal direction and are stacked vertically to form one cage row 50. A plurality of such cage rows 50 are installed in the poultry house.

  A feeding device 52 and an egg collecting device 53 are horizontally provided on the front surface of each cage 51. Chickens in the cage 51 are fed by a feeding device 52. The floor surface of the cage 51 has an inclination that descends from the rear toward the front. Eggs laid off by the chicken roll forward due to the inclination of the floor surface, are discharged from the cage 51, and are collected by the egg collector 53.

  The traveling robot 10 has a traveling carriage 12. The traveling carriage 12 travels in a passage in front of the cage row 50. A pole 15 is erected on the traveling carriage 12, and a camera 13 is fixed to the pole 15. While the traveling carriage 12 travels along the passage of the poultry house, the camera 13 photographs the chicken in the cage 51.

  The camera 13 may be any camera as long as it can acquire images of chickens, and various cameras can be employed. A general visible light color camera is preferably used as the camera 13. In this case, the monitoring image is digital data composed of color information of each pixel.

  The traveling carriage 12 is preferably mounted with the same number of cameras 13 as the number of stages of cages 51 constituting the cage row 50. For example, when the cage row 50 has three stages, three cameras 13 are mounted on the traveling carriage 12. The height of each camera 13 is adjusted in accordance with the corresponding stage cage 51.

  The number of cameras 13 mounted on the traveling carriage 12 may be less than the number of stages of the cage row 50, for example, one. In this case, if the traveling carriage 12 is provided with a mechanism for moving the camera 13 up and down, the entire plurality of cages 51 can be imaged with a small number of cameras 13.

  The feeding device 52 and the egg collection device 53 are provided at an interval in the vertical direction. From between the feeding device 52 and the egg collection device 53, the leg of the chicken can be seen mainly. Hereinafter, the opening between the feeding device 52 and the egg collecting device 53 of the cage 51 is referred to as a “leg opening”. It is preferable to adjust the height and angle of the camera 13 so as to photograph the chicken leg from the leg opening.

  In general, since the inside of a poultry house is dark, the amount of light may be insufficient for photographing with the camera 13. In this case, the illumination 16 may be mounted on the traveling carriage 12. The illumination 16 is fixed to the pole 15. Further, it is preferable to mount the same number of lights 16 as the camera 13 on the traveling carriage 12. In this case, one illumination 16 is provided so as to illuminate the photographing range of one camera 13.

  The illumination 16 only needs to illuminate the photographing range of the camera 13, and various illuminations can be employed. However, it is preferable to use the illumination 16 having a light color temperature of 4,000 to 7,000K. This type of illumination includes illumination commonly referred to as white (4,200 K), day white (5,000 K), and daylight color (6,000 K). Among these, it is preferable to employ daylight white illumination.

  If the color temperature is 4,000 to 7,000 K, the color of light is close to white. If the illumination 16 which emits such light is used, the stress given to a chicken can be made small. Since chickens are accustomed to natural light, it is presumed that the stress applied to chickens can be reduced by using light close to white for lighting. In fact, when red lighting is used, it has been confirmed that chickens make noise and give great stress to chickens. Giving stress to chickens is not preferable because it may adversely affect egg laying.

  Moreover, since the color of the light of the illumination 16 is close to white, the original color of the chicken is easily reflected in the monitoring image photographed by the camera 13. Therefore, as described later, it is suitable for extracting the characteristics of the dead chicken from the color of the monitoring image.

  It is preferable that the lighting 16 is not flashed and is always turned on at least while the patrol robot 10 patrols the poultry house. If it does so, the stress given to a chicken can be reduced. Actually, it has been confirmed that when the illumination 16 is flashed, the chicken makes a noise and gives a great stress to the chicken.

  As shown in FIG. 3, a plurality of cage rows 50 are generally arranged side by side in a poultry house. A passage is provided between the cage rows 50. The traveling robot 10 travels along the route R defined in the passage of the poultry house. That is, the traveling carriage 12 on which the camera 13 is mounted travels along the cage 51 (longitudinal direction of the cage row 50). For this reason, the entire plurality of cages 51 provided in the poultry house can be monitored with a small number of cameras 13 mounted on the traveling carriage 12.

  As the traveling carriage 12, one that travels along a magnetic tape laid on the route R can be used. A traveling vehicle 12 that travels while grasping the surrounding space with a radar such as a laser radar may be used.

  The traveling robot controller 22 starts a traveling instruction to the traveling carriage 12 when a predetermined traveling time has arrived. For example, the traveling robot control device 22 causes the traveling carriage 12 to travel along the route R for a predetermined distance. Next, the traveling robot control device 22 stops the traveling carriage 12 and performs photographing with the camera 13. Thereafter, the traveling robot control device 22 causes the traveling carriage 12 to travel again along the route R for a predetermined distance. By repeating this, the cage 51 is photographed at a plurality of photographing points on the route R. Thereby, monitoring images are acquired for all the cages 51.

  Each cage 51 installed in the poultry house has its position determined by the column number indicating the cage column 50, the column number indicating the upper, middle, lower, etc., the row number allocated along the longitudinal direction of the cage column 50, etc. Can be uniquely identified. The traveling robot control device 22 has information indicating the position of the traveling robot 10 on the route R. From this information, the traveling robot control device 22 can generate position information of the cage 51 shown in the monitoring image photographed by the camera 13. Here, the position information of the cage 51 includes, for example, a column number, a stage number, and a row number.

  Each time the monitoring image is taken by the camera 13, it is transmitted to the chicken life / death determining apparatus 30. In addition, after the patrol robot 10 finishes patrol of the poultry house, for example, when reaching the standby position provided on the route R, a plurality of monitoring images taken in the patrol are transmitted to the chicken life / death determining apparatus 30. Also good.

(Chicken life / death judgment device)
Next, details of the chicken life / death determining apparatus 30 will be described.
FIG. 4 is a block diagram showing the configuration of the chicken life / death determining apparatus 30. The chicken life / death determining apparatus 30 is realized by a computer. The computer may be a dedicated machine or a general-purpose machine. The computer includes a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a main storage device, an auxiliary storage device, and the like.

  A chicken life / death determination program is stored in the storage device of the chicken life / death determination apparatus 30. The chicken life / death determination program may be composed of a plurality of subprograms. The chicken life / death determination program is installed in a computer constituting the chicken life / death determination apparatus 30. When the computer executes the chicken life / death determination program, the function as the chicken life / death determination apparatus 30 is realized. Specifically, an image acquisition unit 31, a position acquisition unit 32, a learning device 33, a life / death determination unit 34, a history storage unit 35, and a notification unit 36 are realized.

  The chicken life / death determination program includes a learned model. When the computer executes the chicken life / death determination program, a learning device 33 is configured on the computer. As an algorithm of the learning device 33, an algorithm suitable for handling a classification problem such as a neural network or a support vector machine is used. The learning device 33 is previously subjected to machine learning. The machine learning method is not particularly limited, but supervised learning is preferably used. When the learning device 33 is a neural network, the weighting coefficient between neurons is optimized by machine learning.

Machine learning is performed, for example, by the following procedure. This procedure is based on supervised learning.
First, learning data used for machine learning is prepared. Learning data consists of training data and evaluation data. The training data includes a plurality of survival images, a plurality of death images, and life / death information of these images. The evaluation data is composed of a plurality of survival images and a plurality of death images, and no life / death information is attached thereto.

  Here, the life-and-death image and the death image are images obtained by photographing chickens in the cage 51 under the same conditions as the camera 13 mounted on the traveling robot 10. The live / dead image includes only live chickens and does not include dead chickens. Death images include dead chickens. Live images may be included in the death image. Life / death information is information indicating whether or not a dead chicken is included in an image. Survival information is attached to the survival image constituting the training data, and death information is attached to the death image. Life / death information is a teacher signal in supervised learning.

  Training data is input to the learning device 33, and feature amounts and patterns of the survival image and the death image are learned. Next, evaluation data is input to the learning device 33 to evaluate whether the output life / death information is correct. The above training and evaluation are repeated until correct life / death information is output for the evaluation data. Thereby, when the learning device 33 is a neural network, the weighting coefficient between neurons is optimized.

  As described above, the learning device 33 performs machine learning in advance using learning data including a survival image and a death image. When the monitoring image captured by the camera 13 is input, the learning device 33 outputs a determination result as to whether or not a dead chicken is included in the monitoring image. Since the learning device 33 performs machine learning using learning data including a survival image and a death image, it can accurately determine whether or not a dead chicken is included in the monitoring image. As a result, it is possible to accurately determine whether a chicken is alive or dead.

  The accuracy of the determination by the learning device 33 depends on the quality of the learning data. For example, there are various aspects of dead chickens. If a dead chicken of a specific aspect is not learned, when a monitoring image including the dead chicken of that aspect is input, life / death may not be accurately determined. Therefore, the inventor of the present application further classifies the survival image and the death image, and causes the learning device 33 to learn various types of chickens, thereby further improving the determination accuracy of the learning device 33.

  Specifically, the living image includes a standing image and a sitting image. As illustrated in FIG. 7, the standing image is an image in which all chickens in the image are standing. Only chicken legs can be recognized mainly from the leg openings of the cage 51. As illustrated in FIG. 8, the sitting image is an image in which some chickens in the image are standing and the remaining chickens are sitting. The chicken body can be recognized from the leg opening of the cage 51.

  The death image includes a fallen image, a missing image, and a discolored image. As illustrated in FIG. 9, the fallen image is an image in which some chickens in the image have fallen without being lost or discolored. From the leg opening of the cage 51, the chicken body which is turned sideways can be recognized. As illustrated in FIG. 10, the defect image is an image in which some chickens in the image have a defect. Here, the deficiency means that a part of the chicken body is missing. This includes the case where the neck of the chicken is missing as shown in FIG. 10A and the case where the chicken wings are peeled off as shown in FIG. A defect occurs when a dead chicken is kicked by another chicken. As illustrated in FIG. 11, the discolored image is an image in which some chickens in the image are discolored. The case where a part of a chicken body is discolored as shown in FIG. 11A and the case where it is discolored due to blood adhesion as shown in FIG. 11B are included. When a dead chicken rots, part of the body changes color, or when a dead chicken is kicked by another chicken and bleeds, blood adheres.

  Note that only one type or a plurality of types of fallen images, missing images, and discolored images may be included as death images. From the viewpoint of improving the determination accuracy of the death image, it is preferable that all these types of images are included in the death image. Moreover, the image containing the dead chicken of another aspect may be contained as a death image.

  As a survival image and a death image used as learning data, an image captured by a camera may be used as it is, or an image obtained by performing some processing on an image captured by a camera may be used. When the processed image is used as the learning data, the same processing is performed on the monitoring image to be determined acquired from the camera 13 and input to the learning device 33.

  As described above, the learning device 33 performs machine learning using learning data including a sitting image as well as a standing image as a living image. Therefore, the learning device 33 stands when all the chickens in the monitoring image are standing (corresponding to the standing image), or some chickens in the monitoring image are standing and the remaining chickens are sitting (sitting image). In the case where the dead chicken is not included in the monitoring image. Even if all chickens in the surveillance image are standing or some chickens are sitting, it is determined that the surveillance image does not include dead chickens. The frequency of determination can be reduced.

  The learning device 33 performs machine learning using learning data including a fallen image having a characteristic of a dead chicken, a missing image, and a discoloration image as a death image. Therefore, the learning device 33 has a case where a part of chickens in the monitoring image has fallen (corresponding to a fallen image), a part of chickens in the monitoring image has a defect (corresponds to a defective image), or When some chickens in the monitoring image are discolored (corresponding to the discolored image), it is determined that a dead chicken is included in the monitoring image. When the chicken in the surveillance image has fallen, missing, or discoloration characteristic of a dead chicken, it is judged that the surveillance image contains a dead chicken, reducing the frequency of misjudging a dead chicken as a live chicken it can.

  When the monitoring image is input, the learning device 33 outputs the life / death determination result of the monitoring image. The life / death determination result is, for example, the probability (probability) that a dead chicken is included in the monitoring image.

  As shown in FIG. 4, the chicken life / death determination apparatus 30 includes an image acquisition unit 31, a position acquisition unit 32, a life / death determination unit 34, a history storage unit 35, and a notification unit 36 in addition to the learning device 33.

  The image acquisition unit 31 acquires a monitoring image from the camera 13 of the traveling robot 10 and inputs the acquired monitoring image to the learning device 33. The image acquisition unit 31 may input the monitoring image acquired from the camera 13 to the learning device 33 as it is, or may input the monitoring image after performing some processing on the monitoring image.

  The position acquisition unit 32 acquires position information of the cage 51 shown in the monitoring image acquired by the image acquisition unit 31. The position acquisition unit 32 acquires the position information of the cage 51 from, for example, the traveling robot control device 22.

  The life / death determination unit 34 performs a final life / death determination based on the life / death determination result output from the learning device 33. The learning device 33 outputs a probability that a dead chicken is included in the monitoring image (hereinafter referred to as “death probability”) as a result of life / death determination. The life / death determination unit 34 determines “death” when the death probability is equal to or higher than the death threshold. The life / death determining unit 34 determines “survival” when the death probability is equal to or less than the survival threshold. Furthermore, the life / death determination unit 34 determines “life / death unknown” when the death probability exceeds the survival threshold and is less than the death threshold.

  The death threshold and the survival threshold are set in advance. The death threshold is higher than the survival threshold. For example, the death threshold is set to 80% and the survival threshold is set to 20%. In this case, if the death probability is 80% or more, it is determined as “death”. If the death probability is 20% or less, it is determined as “survival”. If the probability of death exceeds 20% and is less than 80%, it is determined that “life and death is unknown”.

  In this way, by determining the life and death of chickens in three stages of death, survival, and unknown life and death, it is possible to avoid forced determination of death and survival based on a monitoring image in which it is difficult to determine whether or not life is dead. Therefore, the frequency of erroneous detection can be reduced.

  The history storage unit 35 stores the life / death determination result (death, survival, life / death unknown) by the life / death determination unit 34 in a state in which the result is associated with the position information of the cage 51 where the monitoring image that is the basis of the determination is captured. To do. Here, the position information of the cage 51 is input from the position acquisition unit 32. The traveling robot 10 repeatedly visits the poultry house regularly or irregularly, and takes a monitoring image. The history storage unit 35 stores the determination result as a history for each patrol of the patrol robot 10.

  The notification unit 36 notifies dead chicken information when the life / death determination unit 34 determines death. Notification of dead chicken information is performed, for example, by sending an e-mail to the terminal 40 of the staff. Note that a dedicated application may be installed in the terminal 40 and the dead chicken information may be notified via the application. By notifying dead chicken information, staff can be encouraged to remove dead chickens.

  You may include the monitoring image used as the basis of the death determination in dead chicken information. By notifying the monitoring image, it is possible to determine whether a chicken is alive or dead. Even if the chicken life / death determining apparatus 30 determines that the chicken is dead, if the dead chicken cannot be confirmed in the monitoring image with human eyes, it is not necessary to perform the removal of the dead chicken. Therefore, the burden on the staff can be reduced.

  The position information of the cage 51 where the monitoring image that is the basis for the death determination is taken may be included in the dead chicken information. Since the position information of the cage 51 is notified, the staff can grasp the position of the dead chicken, and the removal work of the dead chicken becomes easy. Moreover, you may include the imaging | photography date of a monitoring image in dead chicken information. Then, the date and time of occurrence of dead chicken can be grasped.

(Processing of dead chicken detection system)
Next, the operation and processing of the dead chicken detection system 1 will be described based on the flowchart shown in FIG.
The patrol robot 10 starts patrol of the poultry house when a predetermined patrol start time arrives. The patrol start time may be, for example, one time of a day or a plurality of times.

  The patrol start time is stored in the patrol robot controller 22. The traveling robot controller 22 gives a traveling instruction to the traveling robot 10 after the traveling start time has arrived (step S21). The traveling robot 10 that has received the travel instruction travels a predetermined distance along the route R set in the poultry house, and then stops (step S31).

  Next, the traveling robot control device 22 issues a photographing instruction to the traveling robot 10 (step S22). The traveling robot 10 that has received the shooting instruction takes a monitoring image with the mounted camera 13 (step S32). Then, the traveling robot 10 transmits the monitoring image to the chicken life / death determining apparatus 30 (step S33).

  The traveling and photographing of the patrol robot 10 are repeated until the patrol robot 10 has patrols the entire poultry house.

  The image acquisition unit 31 of the chicken life / death determining apparatus 30 receives a monitoring image from the traveling robot 10 (step S11). Then, the position acquisition unit 32 of the chicken life / death determining apparatus 30 inquires of the traveling robot control apparatus 22 about the position of the cage 51 where the monitoring image is taken. The traveling robot control device 22 transmits the position information of the cage 51 to the chicken life / death determining device 30 (step S23). Then, the chicken life / death determining apparatus 30 receives the position information of the cage 51 (step S12). Thereby, the chicken life / death determination apparatus 30 will hold | maintain the monitoring image and the positional information corresponding to it. Note that the position acquisition unit 32 may acquire the shooting date and time of the monitoring image from the traveling robot control device 22.

  Next, the chicken life / death determining apparatus 30 performs a life / death determination process for determining the life / death of the chicken based on the acquired monitoring image (step S13). The chicken life / death determining apparatus 30 performs one of survival determination, death determination, and life / death unknown determination. Details of the life / death determination process will be described later.

  When the result of the life / death determination process is a death determination (Yes in Step S14), the notification unit 36 of the chicken life / death determination apparatus 30 notifies the dead chicken information (Step S15). For example, the notification unit 36 transmits an e-mail including dead chicken information to the terminal 40 owned by the staff.

  When the result of the life / death determination process is a survival determination or a life / death unknown determination (No in step S14), the chicken life / death determination apparatus 30 ends the process without performing notification.

  The process from the monitoring image reception (step S11) to the notification (step S15) of the chicken life / death determination apparatus 30 is performed every time a monitoring image is received from the traveling robot 10.

Next, details of the life / death determination process will be described based on the flowchart shown in FIG.
The image acquisition unit 31 of the chicken life / death determining apparatus 30 inputs the monitoring image acquired from the traveling robot 10 to the learning device 33 as it is or after performing some processing. The learning device 33 obtains a probability that a dead chicken is included in the input monitoring image, that is, a death probability (step S13.1).

  The life / death determining unit 34 compares the obtained probability of death with the survival threshold (step S13.2). When the death probability is equal to or less than the survival threshold, the life / death determining unit 34 performs the survival determination (step S13.3). The history storage unit 35 stores information such as the shooting date and time of the monitoring image, the position information of the cage 51, the monitoring image, and the determination result (survival determination).

  When the death probability exceeds the survival threshold value in step S13.2, the life / death determination unit 34 compares the death probability with the death threshold value (step S13.4). When the death probability is equal to or higher than the death threshold, the life / death determination unit 34 performs death determination (step S13.5). The history storage unit 35 stores information such as the shooting date and time of the monitoring image, the position information of the cage 51, the monitoring image, and the determination result (death determination).

  When the death probability is less than the death threshold value in step S13.4, the life / death determination unit 34 performs the life / death unknown determination (step S13.6). The history storage unit 35 stores information such as the shooting date and time of the monitoring image, the position information of the cage 51, the monitoring image, and the determination result (life / death unknown determination).

  As described above, by determining whether a chicken is alive or not in three stages of dying, survival, and unknown mortality, it is possible to avoid forcibly deciding whether it is dying or alive based on a monitoring image that is difficult to determine whether alive or dead. Therefore, the frequency of erroneous detection can be reduced.

  By the way, the notification in step S15 (see FIG. 5) may be performed not only in the case of the death determination but also in the case where the life / death unknown determination continues. For example, the notification unit 36 refers to the determination result history stored in the history storage unit 35, and based on a plurality of monitoring images obtained by repeatedly photographing the cage 51 at the same position, the notification unit 36 determines whether the life / death is unknown. If there is, notify. Here, the continuous number of life / death unknown determinations to be notified may be set to an arbitrary number such as 2 times or 3 times.

  In this case, the notification unit 36 transmits life / death unknown information to the terminal 40. It is preferable that the life / death unknown information includes a monitoring image that is the basis for the life / death unknown determination. By notifying the monitoring image, it is possible to determine whether a chicken is alive or dead. The staff member who received the notification will make a decision on whether the chicken is alive or not, and if necessary (if it is determined that it has died), it may be performed to remove the dead chicken.

  In this way, based on a plurality of time-sequential monitoring images obtained by repeatedly photographing the cage 51 at the same position, notification is made even when it is determined that all are alive / dead. The possibility that dead chickens are left unattended can be reduced.

[Second Embodiment]
Next, a dead chicken detection system according to a second embodiment of the present invention will be described.
As described above, the accuracy of the determination by the learning device 33 depends on the quality of the learning data. The determination accuracy of the learning device 33 is higher when images of various patterns are prepared as death images constituting the learning data.

  As indices for patterning of death images, (1) type of dead chicken, (2) position in cage of dead chicken, (3) angle of dead chicken, (4) orientation of dead chicken, (5) configuration of cage Is mentioned. Hereinafter, it demonstrates in order.

(1) Types of dead chickens Types of dead chickens classify the modes of dead chickens. There are three types of dead chickens: “Fall”, “Defect”, and “Discoloration”. “Falling” means that the chicken has fallen without loss and discoloration. “Deficient” means that a part of the chicken's body is missing. “Discoloration” means that the chicken's body is discolored due to decay, blood adhesion, or the like. The “falling image”, “defect image”, and “discolored image” in the first embodiment correspond to three patterns of images of dead chickens. Note that some of the three patterns of “falling”, “missing”, and “discoloration” may be used, or other modes may be added.

(2) Position of dead chicken in cage The position of dead chicken in the cage classifies the position of dead chicken in the cage. For example, as shown in FIG. 12, with the camera 13 as the front, the inside of the cage 51 is divided into left, middle, and right in the horizontal direction, and divided into front, middle, and back in the depth direction. Then, the inside of the cage 51 is divided into nine regions. In the example shown in FIG. 12, the dead chicken exists in the area on the right back. Nine patterns of images are obtained depending on the difference in the area where the dead chicken exists. The inside of the cage 51 may be divided more finely or more roughly.

(3) Angle of the dead chicken The angle of the dead chicken is a classification of the head direction of the dead chicken when the dead chicken lying on the floor of the cage is viewed from above. For example, as shown in FIG. 13 (A), the case where the head of the dead chicken faces the back direction of the cage 51 is defined as 0 °. Then, as shown in FIG. 13B, when the head of the dead chicken is facing the front of the cage 51, the angle is 180 °. Dividing 360 ° in all directions in the horizontal plane into a range of 15 ° and shooting with a dead chicken facing each direction, 24 patterns of images can be obtained. The angle may be divided more finely or more roughly.

(4) Direction of dead chickens The direction of dead chickens is a classification of angles around the central axis of dead chickens. For example, a dead chicken having the posture shown in FIG. Moreover, let the dead chicken of the attitude | position shown in FIG.14 (B) face right. Thus, there are two patterns of “dead” and “right” as the direction of the dead chicken. Moreover, you may add two patterns, "the prone face" and "the supine".

(5) Cage configuration The cage configuration is a classification of differences in the cage configuration reflected in the death image. Since the cage row 50 is supported by the frame, the frame may or may not appear in the death image depending on the position (column, stage, row) of the cage 51. For example, there may be two patterns with and without frame reflection.

  Of the above (1) Type of dead chicken, (2) Position of dead chicken in cage, (3) Angle of dead chicken, (4) Direction of dead chicken, (5) Cage composition, only one index May be adopted, or a plurality of arbitrary indices may be combined. For example, (1) 3 types of dead chickens, (2) 9 patterns of dead chicken cage positions, (3) 24 dead bird angles, (4) 2 orientations of dead chickens, ( 5) If the cage configuration is two patterns and all of them are combined, there are 2,592 patterns.

  According to the selected index, images are taken while changing the posture of the dead chicken, and a plurality of patterns of images are obtained. Machine learning is performed using these images as death images. By performing machine learning using the death images of various patterns, the erroneous determination rate of the learning device 33 can be reduced.

[Third Embodiment]
Next, a dead chicken detection system according to a third embodiment of the present invention will be described.
The life / death determination unit 34 (see FIG. 4) constituting the chicken life / death determination apparatus 30 may re-determine that it is dead when the life / death unknown determination continues. Specifically, the life / death determination unit 34 refers to a history of determination results stored in the history storage unit 35, and based on a plurality of monitoring images obtained by repeatedly photographing the cage 51 at the same position, all of them. If it is determined that the life / death is unknown, the determination result is changed to “death”. By performing such a determination, it is possible to reduce detection omissions of dead chickens.

Based on the flowchart shown in FIG. 15, the detail of the life-and-death determination process in this embodiment is demonstrated.
Steps S13.1 to S13.5 are the same as those in the first embodiment, and a description thereof will be omitted (see FIG. 6). When the death probability is less than the death threshold value in step S13.4, the life / death determination unit 34 temporarily determines that life / death is unknown (step S13.7). Next, the life / death determination unit 34 acquires the determination result of the cage 51 at the same position in the previous tour from the history storage unit 35 (step S13.8). When the previous determination result is the survival determination or the death determination (No in Step S13.9), the life / death determination unit 34 determines that the life / death is unknown (Step S13.10).

  If the previous determination result acquired in step S13.8 is a life / death unknown determination, that is, if the determination result of the cage 51 at the same position is unknown twice in a row (in the case of Yes in step S13.9), The life / death determining unit 34 re-determines that it is dead (step S13.5). In step S13.9, the number of consecutive life / death unknown determinations may not be two, but can be set to an arbitrary number of three or more.

  As described above, when it is determined that the life / death is unknown based on a plurality of time-sequential monitoring images obtained by repeatedly photographing the cage 51 at the same position, the death is finally determined. . By setting it as such a process, the detection omission of a dead chicken can be reduced.

[Fourth Embodiment]
Below, the dead chicken detection system 4 which concerns on 4th Embodiment of this invention is demonstrated.
As shown in FIG. 16, the dead chicken detection system 4 of the present embodiment includes a thermography camera 17 in addition to a camera 13 (color camera). Since the preliminary configuration is the same as that of the first embodiment, the same members are denoted by the same reference numerals and the description thereof is omitted.

  The thermography camera 17 is a camera that captures the temperature distribution of an object as an image. Similarly to the camera 13, the thermography camera 17 photographs the chicken in the cage 51. Hereinafter, an image obtained by photographing the chicken in the cage 51 by the thermography camera 17 is referred to as a “temperature image”.

  It is preferable that the angle of view of the thermographic camera 17 and the angle of view of the camera 13 are substantially the same. By doing so, it is easy to compare the monitoring image obtained by the camera 13 and the temperature image obtained by the thermography camera 17. Therefore, the thermographic camera 17 is mounted on the traveling robot 10. For example, the thermographic camera 17 is fixed side by side with the camera 13 on the pole 15 of the traveling robot 10 (see FIG. 2).

  The thermographic camera 17 is connected to the in-flight network 11. Therefore, the thermography camera 17 and the chicken life / death determining apparatus 30 can communicate bidirectionally. The temperature image obtained by the thermography camera 17 is transmitted to the chicken life / death determining apparatus 30 via the networks 11 and 21.

  As shown in FIG. 17, the chicken life / death determining apparatus 30 of the present embodiment includes a temperature determining unit 37. The temperature determination part 37 is implement | achieved because a computer runs the chicken life / death determination program. Since the preliminary configuration is the same as that of the first embodiment, the same members are denoted by the same reference numerals, and description thereof is omitted.

  The image acquisition unit 31 acquires a temperature image from the thermography camera 17 and inputs the acquired temperature image to the temperature determination unit 37. When the life / death determination unit 34 determines that the life / death is unknown, the temperature determination unit 37 performs life / death determination again based on the temperature image.

Next, processing of the dead chicken detection system 4 will be described based on the flowchart shown in FIG.
Since the process of the dead chicken detection system 4 of this embodiment is basically the same as that of the first embodiment, differences from the first embodiment will be mainly described.

  The traveling robot 10 that has received the imaging instruction from the traveling robot control device 22 captures a monitoring image with the camera 13 and also captures a temperature image with the thermography camera 17 (step S32). That is, the monitoring image and the temperature image are taken at the same timing for the cage 51 at the same position. The traveling robot 10 transmits the monitoring image and the temperature image to the chicken life / death determining apparatus 30 (step S33).

  The chicken life / death determination apparatus 30 performs a life / death determination process based on the acquired monitoring image and temperature image (step S13).

Based on the flowchart shown in FIG. 18, the detail of the life-and-death determination process in this embodiment is demonstrated.
Steps S13.1 to S13.5 are the same as those in the first embodiment, and a description thereof will be omitted (see FIG. 6). When the death probability is less than the death threshold value in step S13.4, the life / death determination unit 34 temporarily determines that life / death is unknown (step S13.11).

  Next, the temperature determination part 37 acquires the temperature of the chicken whose life is unknown from the temperature image (step S13.12). Specifically, in step S13.1, the learning device 33 obtains the probability (death probability) that the monitoring image includes a dead chicken, and outputs the region in the monitoring image of the chicken that is the basis for calculating the death probability. To do. Therefore, when it is tentatively determined that the life / death is unknown in step S13.11, the region of the chicken whose life / death is unknown in the monitoring image is known from the output of the learning device 33.

  The temperature determination unit 37 acquires from the image acquisition unit 31 a temperature image taken at the same timing as the monitoring image for which the learning device 33 calculates the death probability. For example, the temperature image shown in FIG. 19B is acquired for the monitoring image shown in FIG.

  The temperature determination part 37 demarcates the area | region of the mortality unknown in a temperature image based on the area | region of the mortality unknown in a monitoring image. If the angle of view is approximately the same between the monitoring image and the temperature image, the area in the monitoring image can be replaced with the area in the temperature image as it is.

  Next, the temperature determination part 37 acquires the temperature (chicken temperature) of the area | region of a chicken whose life / death is unknown from a temperature image. Here, the average temperature of the chicken region whose life and death are unknown may be the chicken temperature, and the highest temperature may be the chicken temperature.

  Next, the life / death determining unit 34 compares the chicken temperature obtained by the temperature determining unit 37 with the temperature threshold (step S13.13). Then, when the chicken temperature is equal to or lower than the temperature threshold, it is determined again as “dead” (step S13.3). Conversely, when the chicken temperature exceeds the temperature threshold, it may be determined again as “survival” (step S13.4).

  Here, the temperature threshold is set to a temperature (for example, 30 ° C.) lower than the average body temperature of the live chicken. Chickens begin to cool immediately after they die and the temperature drops. If the temperature of an undetermined chicken is lower than the temperature threshold, it is likely that it is a dead chicken.

  Note that the temperature threshold value may be determined in two stages, and if the chicken temperature is between the two threshold values, it may be determined that the life / death is unknown.

  As described above, when it is determined that the life / death is unknown based on the monitoring image, the life / death determination is performed again based on the temperature image. Therefore, the number of cases processed as life / death unknown can be reduced. Therefore, it is possible to reduce the work for the staff to judge whether the chicken is alive or dead by looking at the monitoring image. Moreover, the detection omission of a dead chicken can be reduced.

[Other Embodiments]
The chicken life / death determining apparatus 30 in the above embodiment is configured to perform the determination of the chicken's life / death based on the monitoring image using the learned learning device 33, but is not limited thereto. The chicken life / death determination apparatus 30 only needs to perform some processing on the monitoring image to extract features and determine the life / death of the chicken. By determining whether a chicken is live or not based on whether or not a dead chicken is included in the monitoring image, it is possible to accurately detect a dead chicken.

  Specifically, it may be determined that the chicken is alive when all the chickens in the monitoring image are standing, or when some of the chickens in the monitoring image are standing and the remaining chickens are sitting. By making such a determination, it is possible to reduce the frequency of erroneously determining a sitting chicken as a dead chicken.

  Also, if some chickens in the surveillance image have fallen, some chickens in the surveillance image have defects, or some chickens in the surveillance image have discolored, it is determined to be dead. That's fine. When the chicken in the monitoring image has a fall, defect, or discoloration that is a characteristic of the dead chicken, by determining that the chicken is dead, the frequency of misjudging the dead chicken as a live chicken can be reduced.

  The camera 13 and the thermographic camera 17 that photograph the chicken in the cage 51 may not be mounted on the traveling carriage 12. For example, the camera 13 and the thermographic camera 17 may be provided for each cage 51. Similarly, the illumination 16 may not be mounted on the traveling carriage 12. The illumination 16 may be provided directly on the cage 51, or the illumination 16 that illuminates the entire poultry house may be used.

  A display object such as a plate indicating position information of the cage 51 may be directly attached to the cage 51. If the display object is stored in the imaging region and the monitoring image is captured, the position information of the cage 51 can be grasped by confirming the monitoring image notified by the staff. Therefore, in this case, the chicken life / death determining apparatus 30 may not acquire the position information of the cage 51 from the traveling robot control apparatus 22.

  Notification of dead chicken information is not limited to e-mail. The method of displaying on the display etc. which were connected to the chicken life / death determination apparatus 30 may be used.

  The installation position of the chicken life / death determining apparatus 30 is not particularly limited. You may install in a chicken farm and may install outside a chicken farm. When the chicken life / death determining apparatus 30 is installed outside the chicken farm, the chicken life / death determining apparatus 30 communicates with the traveling robot 10 via the Internet, for example. In this case, you may comprise the chicken life / death determination apparatus 30 as what is called a cloud system. Further, the chicken life / death determining apparatus 30 may be mounted on the traveling robot 10.

Next, examples will be described.
Example 1
A learned model was created by machine learning of a neural network constructed on a computer by supervised learning. The learning data consists of a survival image and a death image. As survival images, 700 standing images and 700 sitting images were prepared. As death images, 72 fallen images, 5 missing images, and 14 discolored images were prepared.

  877 new survival images that are not used as learning data are input to the learned model after learning using the learning data, and life / death determination is performed for each. Here, if the death probability output by the learned model is 80% or more, it is determined as “death”. If the death probability is 20% or less, it is determined as “survival”. If the probability of death exceeds 20% and is less than 80%, it is determined that “life and death is unknown”.

  As a result, the probability of correctly determining “survival” was 79.0%. The probability of determining “life and death unknown” was 14.7%. The probability of misjudging as “dead” was 6.3%.

  In addition, 545 new death images that are not used as learning data are input to the learned model, and life / death determination is performed for each. As a result, the probability of correctly determining “death” was 84.0%. The probability of determining “life and death unknown” was 9.9%. The probability of misjudging as “surviving” was 6.1%.

(Example 2)
A learned model was created by machine learning of a neural network constructed on a computer by supervised learning. The learning data consists of a survival image and a death image. As survival images, 700 standing images and 700 sitting images were prepared.

  As a death image, (1) the type of dead chicken, (2) the position of the dead chicken in the cage, (3) the angle of the dead chicken, (4) the direction of the dead chicken, and (5) An image obtained by photographing was prepared.

  Here, (1) the types of dead chickens were three patterns of “falling”, “missing”, and “discoloration”. (2) The position of the dead chicken in the cage was 9 patterns divided into 3 in the horizontal direction and 3 in the depth direction. (3) The angle of the dead chicken was 24 patterns in which 360 ° in all directions was divided into 15 ° ranges. (4) The direction of the dead chicken was “left-facing” and “right-facing”. (5) The structure of the cage was two patterns with and without a frame reflection. As these combinations, 2,592 pattern images were prepared one by one.

  251 new survival images that are not used as learning data are input to the learned model after learning using the learning data, and life / death determination is performed for each. As a result, the probability of correctly determining “survival” was 83.7%. The probability of determining “life and death unknown” was 15.9%. The probability of misjudging as “dead” was 0.4%.

  In addition, 265 new death images that are not used as learning data are input to the learned models, and life / death determination is performed for each. As a result, the probability of correctly determining “death” was 75.8%. The probability of determining “life and death unknown” was 20.4%. The probability of misjudging as “surviving” was 3.8%.

  Compared to the first embodiment, the second embodiment has a 5.9% lower probability of misjudging the survival image as “dead”. In addition, compared to the first embodiment, the second embodiment has a 2.3% lower probability of misjudging a dead image as “surviving”. From this, it was confirmed that if the learning method of Example 2 is adopted, the erroneous determination rate of the learning device can be reduced.

DESCRIPTION OF SYMBOLS 1 Dead chicken detection system 10 Traveling robot 12 Traveling carriage 13 Camera 16 Illumination 22 Traveling robot control apparatus 30 Chicken life / death determination apparatus 31 Image acquisition part 32 Position acquisition part 33 Learning device 34 Life / death determination part 35 History storage part 36 Notification part 40 Terminal

Claims (26)

A camera to shoot the chicken in the cage,
A dead chicken detection system comprising: a chicken life / death determining apparatus that determines whether a chicken is dead or not based on whether or not a dead chicken is included in a monitoring image acquired from the camera. The apparatus for determining whether a chicken is alive or dead is preliminarily machine-learned using learning data including an image obtained by photographing a chicken in a cage and not including a dead chicken and a death image including a dead chicken. The dead chicken detection system according to claim 1, wherein the learned learning device is configured. The survival image includes a standing image in which all chickens in the image are standing and a sitting image in which some chickens in the image are standing and the remaining chickens are sitting. Item 2. The dead chicken detection system according to item 2. The death image includes a fallen image in which some chickens in the image have fallen, a defective image in which some chickens in the image are defective, and a discolored image in which some chickens in the image are discolored. The dead chicken detection system according to claim 2 or 3, wherein one or more of them are included. The death image was obtained by changing one or more of the types of dead chicken, the position of the dead chicken in the cage, the angle of the dead chicken, the orientation of the dead chicken, and the configuration of the cage. The dead chicken detection system according to claim 2 or 3, wherein a plurality of patterns of images are included. The chicken life / death determining device is:
When the probability that a dead chicken is included in the monitoring image is equal to or higher than a death threshold,
When the probability that a dead chicken is included in the monitoring image is less than or equal to a survival threshold,
The dead chicken detection according to any one of claims 2 to 5, wherein when the probability that a dead chicken is included in the monitoring image exceeds the survival threshold and is less than the death threshold, it is determined that the life or death is unknown. system. The chicken life / death determining apparatus re-determines that the animal is dead when it is determined that the life / death is unknown based on a plurality of the monitoring images obtained by repeatedly photographing the cage at the same position. Item 6. The dead chicken detection system according to Item 6. A thermography camera for photographing the chicken in the cage;
When the life / death determination apparatus determines that the life / death is unknown based on the monitoring image, the chicken / life determination apparatus acquires the temperature of the area of the life / death chicken in the temperature image acquired from the thermography camera, and the temperature is equal to or lower than the temperature threshold. The dead chicken detection system according to claim 6, wherein it is re-determined as dead. The chicken life / death determining device is dead in the monitoring image when all the chickens in the monitoring image are standing or when some chickens in the monitoring image are standing and the remaining chickens are sitting. The dead chicken detection system according to claim 1, wherein it is determined that no chicken is included. The chicken life / death determining device may have some chickens in the monitoring image fall down, some chickens in the monitoring image may be defective, or some chickens in the monitoring image may be discolored. 10. The dead chicken detection system according to claim 1, wherein it is determined that a dead chicken is included in the monitoring image. The dead chicken detection system according to any one of claims 1 to 10, wherein the chicken life / death determining apparatus notifies dead chicken information when it is determined that the chicken is dead. 12. The dead chicken detection system according to claim 11, wherein the dead chicken information includes the monitoring image that is a basis for the death determination. The dead chicken detection system according to claim 11 or 12, wherein the dead chicken information includes position information of the cage where the monitoring image that is a basis for the death determination is taken. The dead chicken detection system according to any one of claims 1 to 13, further comprising an illumination having a color temperature of 4,000 to 7,000K that illuminates a photographing range of the camera. The dead chicken detection system according to claim 14, wherein the illumination is always on. The dead chicken detection system according to any one of claims 1 to 15, further comprising a traveling carriage on which the camera is mounted and travels along the cage. A chicken life / death determination program for causing a computer to function based on a monitoring image obtained by photographing a chicken in a cage to determine the life / death of the chicken,
A learning machine that has been pre-machine-learned using learning data that includes images of live images that do not include dead chickens, and images that are obtained by photographing chickens in cages and that contain dead chickens. And
When the monitoring image is input to the learning device, the computer is caused to function so as to output a determination result as to whether or not a dead chicken is included in the monitoring image. The survival image includes a standing image in which all chickens in the image are standing and a sitting image in which some chickens in the image are standing and the remaining chickens are sitting. Item 18. The life / death determination program for chickens according to Item 17. The death image includes a fallen image in which some chickens in the image have fallen, a defective image in which some chickens in the image are defective, and a discolored image in which some chickens in the image are discolored. Among them, one or more types are included, The chicken life / death determination program according to claim 17 or 18. The death image was obtained by changing one or more of the types of dead chicken, the position of the dead chicken in the cage, the angle of the dead chicken, the orientation of the dead chicken, and the configuration of the cage. The life / death determination program for chickens according to claim 17 or 18, wherein a plurality of patterns of images are included. When the probability that a dead chicken is included in the monitoring image is equal to or higher than a death threshold,
When the probability that a dead chicken is included in the monitoring image is less than or equal to a survival threshold,
The computer is caused to function so as to determine that the life and death are unknown when the probability that a dead chicken is included in the monitoring image exceeds the survival threshold and is less than the death threshold. Chicken life / death determination program described in 1. The computer is caused to function so as to re-determine that it is dead when all of the monitoring images obtained by repeatedly photographing the cage at the same position are determined to be unknown. The chicken life / death determination program according to 21. When it is determined that the life / death is unknown based on the monitoring image, the temperature of the area where the life / death is unknown is acquired from the temperature image obtained by photographing the chicken in the cage with a thermography camera, and the temperature is a temperature threshold value. 22. The chicken life / death determination program according to claim 21, wherein the computer is caused to function so that it is determined again as dead in the following cases. A chicken life / death determination program for causing a computer to function based on a monitoring image obtained by photographing a chicken in a cage to determine the life / death of the chicken,
When all the chickens in the monitoring image are standing, or when some of the chickens in the monitoring image are standing and the remaining chickens are sitting, the computer is caused to function as being determined to be alive. Characteristic chicken life / death determination program. A chicken life / death determination program for causing a computer to function based on a monitoring image obtained by photographing a chicken in a cage to determine the life / death of the chicken,
If some chickens in the monitoring image have fallen, some chickens in the monitoring image have a defect, or some chickens in the monitoring image are discolored, it is determined to be dead A program for determining whether a chicken is alive or not is characterized by causing a computer to function. A chicken life / death determining apparatus comprising a computer in which the chicken life / death determining program according to any one of claims 17 to 25 is installed.