JP7445612B2 - Pig breeding support device, pig breeding support method, and pig breeding support program - Google Patents

Description

The present invention relates to a pig breeding support device, a pig breeding support method, and a pig breeding support program.

There is a known method of estimating the weight of a livestock by irradiating the body surface of the livestock with a slit-shaped pattern of light and analyzing the shape of the bright line (for example, see Patent Document 1).

JP 2019-45478 Publication

When raising pigs, a method is generally adopted in which they are raised in groups in cages or divided areas called pens. Even when pigs are housed in pens and raised in groups under the same environment, individual pigs may exhibit variations in growth. It is desirable for all pigs in the same pen to grow evenly, but for pigs whose growth is delayed, measures are taken such as transferring them to another breeding group that is generally smaller in size. However, in pig farms where each pen houses, for example, about 10 pigs, and where multiple such pens are provided, the work of finding pigs with delayed growth has become a heavy burden on the keepers. Furthermore, it is also difficult in reality to accurately irradiate each animal with a pattern of light and estimate the weight of each animal.

The present invention was made to solve such problems, and it is a method that requires excessive cost and labor to determine whether there are pigs whose growth is delayed among pigs reared in groups in pens. The purpose of the present invention is to provide a pig breeding support device, etc., which can notify the breeder in a timely manner without causing any problems.

A pig rearing support device according to a first aspect of the present invention includes an acquisition unit that acquires image data of an image captured by a camera installed toward a pen in which pigs are reared in groups, and a a detection unit that extracts the buttock of the pig and detects the width of the buttock of the pig based on the extracted buttock image; a calculation unit that calculates a statistical value of the width of the buttock detected by the detection unit during a set period; and a notification section that issues a notification when the detection section detects a hip width that shows a deviation of more than a threshold value from the statistical value.

A pig rearing support device according to a second aspect of the present invention includes an acquisition unit that acquires image data of an image captured by a camera installed facing a pen in which pigs are reared in groups, and a A detection unit that extracts the abdomen of It includes a calculation unit that calculates a statistical value of the detected wide/narrow ratio, and a notification unit that issues a notification when the sensing unit detects a wide/narrow ratio that shows a deviation of a threshold value or more from the statistical value.

A pig rearing support method according to a third aspect of the present invention includes an acquisition step of acquiring image data of an image taken by a camera installed facing a pen in which pigs are reared in groups, and a step of acquiring image data of an image of the pigs from the image data. a detection step that extracts the buttock of the pig and detects the width of the buttock of the pig based on the extracted buttock image; a calculation unit that calculates a statistical value of the buttock width detected by the detection step during a set period; and a notification step of issuing a notification when the detection step detects a hip width that shows a deviation of more than a threshold value from the statistical value.

A pig rearing support method according to a fourth aspect of the present invention includes an acquisition step of acquiring image data of an image taken by a camera installed facing a pen in which pigs are reared in groups, and a step of acquiring image data of an image of the pigs from the image data. A detection step of extracting the abdomen of the patient and detecting the widest/narrowing ratio, which is the ratio of the widest abdominal width to the narrowest abdominal width, based on the extracted abdominal image; The present invention includes a calculation unit that calculates a statistical value of the detected wide-narrow ratio, and a notification step that issues a notification when a wide-narrow ratio that shows a deviation of more than a threshold value from the statistical value is detected in the detection step.

A pig rearing support program according to a fifth aspect of the present invention includes an acquisition step of acquiring image data of an image captured by a camera installed facing a pen where pigs are reared in groups, and a step of acquiring image data of an image of the pigs from the image data. a detection step that extracts the buttock of the pig and detects the width of the buttock of the pig based on the extracted buttock image; a calculation unit that calculates a statistical value of the buttock width detected by the detection step during a set period; The computer is caused to execute a notification step of issuing a notification when the detection step detects a hip width that shows a deviation of a threshold value or more from the statistical value.

A pig rearing support program according to a sixth aspect of the present invention includes an acquisition step of acquiring image data of an image taken by a camera installed facing a pen in which pigs are reared in groups, and a step of acquiring image data of an image of the pigs from the image data. A detection step of extracting the abdomen of the patient and detecting the widest/narrowing ratio, which is the ratio of the widest abdominal width to the narrowest abdominal width, based on the extracted abdominal image; The computer is caused to execute a calculation unit that calculates a statistical value of the detected wide-narrow ratio, and a notification step that issues a notification when a wide-narrow ratio that shows a deviation of more than a threshold value from the statistical value is detected in the detection step.

According to the present invention, a pig breeding support device is capable of informing a breeder in a timely manner whether there is a pig whose growth is delayed among the pigs reared in a group in a pen without requiring excessive cost or labor. etc. can be provided.

1 is a diagram showing an overall picture of a pig farming environment in which a pig breeding support device according to a first example of the present embodiment is adopted. It is a diagram showing the hardware configuration of a pig breeding support device and peripheral devices. FIG. 2 is a diagram illustrating the concept of processing from extracting a pig's buttock from a captured image to generating a buttock image. It is a figure explaining the concept of the process of measuring buttock width. FIG. 3 is a flow diagram illustrating a processing procedure of a calculation unit. It is a figure showing the whole picture of the pig breeding environment which adopted the pig breeding support device concerning the 2nd example of this embodiment. FIG. 3 is a diagram illustrating a method of generating a buttock image when a stereo camera is used. It is a figure showing the whole picture of the pig breeding environment which adopted the pig breeding support device concerning the 3rd example of this embodiment. It is a diagram showing the hardware configuration of a pig breeding support device and peripheral devices. FIG. 3 is a diagram illustrating the difference in abdominal wide/narrow ratio between normal pigs and lean pigs. FIG. 3 is a diagram illustrating a procedure for detecting abdominal wide/narrow ratio using a learning model. FIG. 2 is a flow diagram illustrating a processing procedure of a calculation unit.

The present invention will be described below through embodiments of the invention, but the claimed invention is not limited to the following embodiments. Furthermore, not all of the configurations described in the embodiments are essential as means for solving the problem.

FIG. 1 is a diagram showing an overall picture of a pig farming environment in which a pig breeding support device according to a first example of the present embodiment is adopted. The pig farm includes a plurality of pens 301 separated by walls and fences. A plurality of pigs 302 (for example, about 10 pigs) are housed in each pen 301 and raised in groups. Note that the number of pigs 302 raised in each pen 301 can be adjusted depending on the breed of pigs 302, the breeding environment, etc. The pigs 302 housed in each pen 301 are of similar size at the start of breeding, and are raised in groups until they grow to a size that can be shipped. That is, in principle, the group of pigs 302 is accommodated in and exited from the pen 301 all at once.

The camera unit 210 is installed, for example, suspended from the ceiling so as to be able to take an overhead view of a predetermined position within the pen 301. The camera unit 210 is installed so as to image a range of a viewing angle Fr that is a bird's-eye view of the pig 302 foraging with its head immersed in the feeding container 303 from behind, for example. According to the camera unit 210 installed in this manner, it is possible to obtain an image in which the buttocks of the pig 302 that are foraging are directly facing each other.

The camera unit 210 is not limited to the case where the predetermined position is near the feeding container 303, but if it can be expected to take an image of the buttocks of the pig 302 moving around inside the pen 301 while facing directly, the camera unit 210 may be set at a predetermined position near the feeding container 303. You can set it up as a location. For example, the predetermined position may be near the water supply container 304. The feeding area and the watering area are places that the housed pigs 302 visit in turn, so they are convenient for randomly acquiring images of the buttocks of many pigs 302.

The camera unit 210 converts the captured image into image data and transmits it to the server 100 via the network 200. Specifically, a wireless unit 230 installed within the facility is connected to the network 200, and the camera unit 210 can transmit image data to the server 100 by establishing wireless communication with the wireless unit 230. can. Note that the network 200 that connects the camera unit 210 and the server 100 may use the Internet or an intranet, or if the management facility where the server 100 is installed is located in a pig farm, short-range wireless communication may be used. May be adopted.

A keeper who takes care of pig 302 may have a keeper terminal 220. The zookeeper terminal 220 is, for example, a tablet terminal or a smartphone, and can exchange various information with the server 100 via the wireless unit 230 and the network 200. For example, the breeder can input the breeding record into the breeder terminal 220 and transfer it to the server 100, or can call up the breeding record stored in the server 100. Further, when the zookeeper terminal 220 receives a notification to be described later, the zookeeper terminal 220 displays a notification to that effect.

A server 100 as a pig breeding support device is installed in the management facility. Server 100 is connected to network 200 . The server 100 sequentially acquires image data from the camera unit 210 installed in each pen 301, and detects pigs 302 with growth retardation based on the image data. The server 100 can display detection status and statistical values to be described later on the display monitor 150 in response to a request from an administrator or a caretaker. The display monitor 150 is, for example, a monitor equipped with a liquid crystal panel. Further, the server 100 is connected to an input device 160 that receives operations from a manager or a caretaker. The input device 160 includes a keyboard, a mouse, a touch panel superimposed on the display surface of the display monitor 150, and the like.

When pigs are raised in groups in pens, some pigs may not be able to collect enough food due to competition with other pigs or poor physical condition. These pigs will have slower growth than other pigs. More specifically, their weight increases at a slower pace than other pigs, and after a certain period of time this becomes apparent as a difference in body size. Such stunted pigs cannot be shipped together with other pigs. For this reason, pigs with growth retardation are often moved to pens with younger pigs of similar size to extend their rearing period, or they are kept in individual cages and given special treatment.

However, in relatively large-scale pig farms where multiple pens are set up in a pigpen, it is difficult for keepers to regularly observe a large number of pigs and identify pigs with developmental delays. This work required a great deal of effort and skill from the zookeepers. The pig breeding support device in this embodiment supports breeding work by detecting whether a pig with growth retardation exists in the pen 301 and notifying the breeder.

As shown in FIG. 1, when there are multiple pens 301 in a pig farm, the breeder can simply receive information from the pig breeding support device as to which pen 301 contains pigs with growth retardation. A lot of effort can be saved. In other words, the breeder can reduce the frequency of observing the pen, and when he/she receives the notification, he/she can search for pigs with growth retardation in that specific pen. Even for inexperienced breeders, it is relatively easy to find pigs with developmental delays in the limited number of pigs in one pen.

As a result of repeated research on a method to more easily evaluate the growth rate of pigs housed in pens, the inventor of the present application found that there is a strong correlation between the width of the buttock of a pig and the weight of the pig. I found it. Generally, there is a certain degree of correlation between the size of any body part of a pig and its body weight, but the correlation between rump width and body weight shows little variation between individuals, and The correlation was found to be particularly stronger than that between body part and body weight. It was also found that if a pig's buttock is the subject of observation, it is possible to accurately and easily perform image analysis based on its position and skin shape. Therefore, in this embodiment, it is determined whether or not there are pigs with developmental delays in the breeding group by observing the buttocks of pigs that move randomly within the pen and detecting the width of the buttocks.

FIG. 2 is a diagram showing the hardware configuration of the server 100 and peripheral devices as a pig breeding support device. The server 100 can be connected to the display monitor 150, input device 160, camera unit 210, and zookeeper terminal 220, as described above.

The server 100 mainly includes a calculation section 110, a storage section 120, and a communication unit 130. The calculation unit 110 is a processor (CPU: Central Processing Unit) that controls the server 100 and executes programs. The processor may be configured to cooperate with an arithmetic processing chip such as an ASIC (Application Specific Integrated Circuit) or a GPU (Graphics Processing Unit). The calculation unit 110 reads out the pig rearing support program stored in the storage unit 120 and executes various processes related to support for pig rearing.

The storage unit 120 is a nonvolatile storage medium, and is configured by, for example, an HDD (Hard Disk Drive). In addition to programs for controlling and processing the server 100, the storage unit 120 can store various parameter values, functions, display element data, lookup tables, etc. used for control and calculation. The storage unit 120 particularly stores a detection neural network 121. When the detection neural network 121 receives an image captured by the camera unit 210, the detection neural network 121 identifies whether or not the image is captured by the pig 302 with the buttocks directly facing each other. The details will be described later. Note that the storage unit 120 may be configured with a plurality of pieces of hardware; for example, the storage medium that stores the program and the storage medium that stores the detection neural network 121 may be configured with separate hardware.

The communication unit 130 includes, for example, a LAN unit, and transmits an imaging control signal generated by the calculation unit 110 to the camera unit 210 via the network 200, and sends image data sent from the camera unit 210 to the calculation unit 110. or hand it over. It also relays data exchange executed between the zookeeper terminal 220 and the calculation unit 110. Note that the communication unit 130 can also relay data and control signals to and from other external devices. For example, it can also be used to import update data for the pig breeding support program and the detection neural network 121 from an external server.

The calculation unit 110 also plays a role as a functional calculation unit that executes various calculations according to the processing instructed by the pig breeding support program. The calculation unit 110 can function as an acquisition unit 111, a detection unit 112, a calculation unit 113, and a notification unit 114. The acquisition unit 111 acquires image data of an image captured by the camera unit 210 and delivers it to the detection unit 112. The detection unit 112 extracts the buttock of the pig from the image of the image data received from the acquisition unit 111, and detects the width of the buttock of the pig based on the extracted buttock image. The calculation unit 113 calculates the statistical value of the hip width detected by the detection unit 112 during the set period. The notification unit 114 issues a notification when the detection unit 112 detects a hip width that indicates a deviation of a preset threshold value or more from the statistical value calculated by the calculation unit 113. Each specific calculation process will be described later.

FIG. 3 is a diagram illustrating the concept of processing from extracting a pig's buttock from a captured image to generating a buttock image. The camera unit 210 sequentially images the pigs 302 that alternately visit the feeding container 303, and the acquisition unit 111 acquires image data of the captured images via the network 200. As described above, since the camera unit 210 is installed to capture an image of a specific viewing angle Fr, for example, the pig 302 to be imaged has the same image as long as the buttocks are facing directly. It can be assumed that it is located at a distance.

The calculation unit 110 finds the tail of the pig 302 from the captured image, cuts out the area around it at a preset image size, and generates a cutout image. Specifically, for example, the region of the tail is found from the captured image by performing matching processing with a large number of template images prepared in advance as tail images, and the region of the tail is found from the captured image, and the region of the tail is included, and the body boundary on the right side and the body boundary on the left side are also included. A cutout image is generated by cutting out the image as follows. Then, the number of pixels _DL from the center of the tail to the left body boundary and the number of pixels _DR from the right body boundary are calculated. In the case of FIG. 6A, since the pig 302 is facing diagonally, D _L ≠ D _R , and it can be determined that the buttocks are not facing directly, so this cropped image is not adopted as the buttock image. On the other hand, in the case of FIG. 6(B), since the pig 302 is facing forward, D _L =D _R , and it can be determined that the buttocks are facing directly, so this cropped image is adopted as the buttock image. In other words, an image in which the tail is located at the center of the buttocks in the width direction is adopted as the buttock image. Note that whether D _L and D _R are equal or not is determined by taking into consideration a tolerance set according to the number of pixels in the horizontal width of the cropped image.

The process of generating a buttock image from a captured image can be realized by the image processing described above, but in this embodiment, the detection unit 112 uses the detection neural network 121 to execute this series of processes. The detection neural network 121 uses supervised learning in advance to learn by giving a considerable amount of training data linked to an image in which D _L = D _R among images of the rear view of a pig where the tail is visible, as the correct answer. It was created. The center coordinates of the tail are given to the training data, and the detection neural network 121 that has learned such training data extracts an image of the back of the pig that satisfies D _L = D _R , and also extracts the center coordinates of the tail. Output. The training data is created, for example, by an operator selecting an image of the back of a pig that satisfies D _L =D _R , and determining and linking the center coordinates of the tail. The detection neural network 121, which is a learning model generated in this way, is incorporated into the server 100, which is a pig breeding support device, and is made available for use.

For example, if a pig farm is provided with eight pens 301 to be observed, image data is sequentially sent to the server 100 from eight camera units 210 that provide an bird's-eye view of each pen 301. When the acquisition unit 111 acquires these images, they are delivered to the detection unit 112. The detection unit 112 sequentially inputs the images read out from the storage unit 120 to the detection neural network 121. The detection neural network 121 executes arithmetic processing every time an image is input, and determines whether the image is an image of the back view of a pig that satisfies D _L =D _R . If it is determined that the image is a rear view that satisfies such conditions, the center coordinates of the tail are output. The detection unit 112 generates a buttock image by cutting out the image whose center coordinates have been output by the detection neural network 121 at a preset image size using the center coordinates as a reference.

FIG. 4 is a diagram illustrating the concept of the process of measuring the width of the buttock. The buttock width is calculated from the buttock image. The range for detecting the hip width is defined as a band-shaped range sandwiched between a range of p ₀ pixels from the recognized center position of the tail toward the back and a range of p ₀ pixels toward the feet. The detection unit 112 sequentially counts the number of pixels from the left body boundary to the right body boundary in the horizontal direction within this band-shaped range, and determines the width at the horizontal position where the counted number is maximum as the buttock width. That is, the maximum width in a predetermined range with the center of the tail as the center in the height direction is determined as the buttock width (pixel). If the detection unit 112 is able to detect the width of the buttock from the captured image thus received, it passes the width of the buttock to the calculation unit 113 and the notification unit 114. The calculation unit 113 cumulatively acquires the buttock width from the start of the set observation period, and calculates the statistical value thereof. In this embodiment, the average value and standard deviation are calculated.

When a predetermined number of hip widths are detected in the pen 301 to be observed, the statistical value calculated by the calculation unit 113 can be used to represent the characteristics of the whole pig kept in the pen 301. The prescribed number is set to, for example, 80% of the number of pigs housed in the pen 301. If a value of around 80% is set, it can be expected that even if a single pig is observed multiple times, the effect will be kept small. If the subsequently detected hip width shows a deviation of more than a threshold value from the calculated statistical value, it can be estimated that the pig whose hip width was detected has a growth abnormality.

In this embodiment, if the hip width is less than α×σ with respect to the average value, it is determined that the pig is growth retarded. Here, α is a coefficient and σ is a standard deviation. α is a value set in advance by an administrator or the like, depending on the age and type of pigs housed, the degree of variation in growth to be allowed, etc. For example, if α = 2 and the variation in growth follows a normal distribution, approximately 95.4% of pigs will be judged as normal, approximately 2.3% will be retarded, and approximately 2.3% will be judged as normal. of pigs are judged to be overgrown. In this embodiment, since we are particularly focused on growth retardation, the notification unit 114 sends a notification to the effect that there are pigs with growth retardation when the rump width is less than α×σ with respect to the average value. The information is sent to the employee terminal 220. Note that the statistical value (including deviation) is not limited to the above, and for example, kurtosis or skewness may be taken into consideration, or the median value may be used.

Next, the processing procedure of the pig rearing support method using the server 100 will be explained. FIG. 5 is a flow diagram illustrating the processing procedure of the calculation unit 110. The flow starts from the moment an instruction to start observation is received.

The acquisition unit 111 acquires image data of img _n from the camera unit 210 of the n-th pen via the communication unit 130 in step S101. For example, if n=1, image data of img ₁ is acquired from the camera unit 210 directed toward the first pen. The acquisition unit 111 delivers the acquired image data to the detection unit 112. n is incremented each time step S101 is executed.

In step S102, the detection unit 112 inputs the image img _n of the received image data to the detection neural network 121, as described using FIG. If a buttock image is obtained from the output, the process advances to step S103, and if not, the process returns to step S101.

After proceeding to step S103, the detection unit 112 detects the buttock width from the buttock image, as described using FIG. 4. Once the hip width is detected, the result is delivered to the calculation unit 113 and notification unit 114. When the calculation unit 113 receives the newly detected hip width from the detection unit 112, it updates the statistical value calculated from the hip width received so far. The calculation unit 113 then proceeds to step S105 and checks whether the number of buttock widths received from the time point when observation was started exceeds a specified number. If the number exceeds the specified number, the statistical value calculated in step S104 and information indicating that the specified number has been exceeded are transferred to the notification unit 114, and the process proceeds to step S106. If the number has not been exceeded, the process returns to step S101.

Proceeding to step S106, the notification unit 114 determines that the hip width received from the detection unit 112 is greater than or equal to the preset standard deviation from the statistical value received from the calculation unit 113. Check whether it exists. In this embodiment, it is checked whether the buttock width received from the detection unit 112 is less than the average value by α×σ, and if it is, the process proceeds to step S107; if not, the process proceeds to step S107. Skip and proceed to step S108. If the process proceeds to step S107, the notification unit 114 sends a notification to the keeper terminal 220 that there is a pig that is considered to be developmentally delayed, with the pen information of the detected pen added. The pen information is, for example, a pen number given to identify a pen. Once the notification is issued, the process advances to step S108. In step S108, the calculation unit 110 checks whether the observation period has ended. If the observation period has not ended, the process returns to step S101 to continue observation. If it has been completed, the series of processing is completed.

In addition, in the above processing procedure, judgments are not made for each detected hip width until the number of hip widths exceeds the specified number, that is, until the calculated statistical value can be evaluated as representing the characteristics of the group. Not yet. However, after the predetermined number is exceeded, each of the hip widths accumulated up to that point may be determined. By determining the accumulated rump width in this manner, it is possible to confirm the presence or absence of growth retardation even in pigs 302 imaged at an early stage after the start of observation.

Furthermore, by limiting the observation period to a preset period, changes in statistical values due to developmental progress can be removed. For example, by calculating the average value by mixing the buttock width from one month ago and the current buttock width, there will be no deviation from the current average value. The observation period is set to a period (for example, 3 days) that does not cause such inconvenience. When a new observation period starts, the previous statistical values are reset. Alternatively, it can be used as a past reference value.

Next, a pig breeding support device according to a second example of this embodiment will be described. In the first embodiment, the camera unit 210 is set to take an image of the feeding container 303 and the like at a relatively narrow angle of view Fr. Furthermore, since the camera unit 210 outputs a two-dimensional image, the width (pixels) of the buttock determined from the buttock image was detected on the assumption that the distance from the camera unit 210 to the buttock of the pig was a constant distance. . However, if the camera unit 210 is changed to a camera unit that can output distance images, it is possible to extract more buttocks images without being restricted by these assumptions. Although a stereo camera, a TOF camera, or the like may be used as the camera unit that outputs a distance image, a case where a stereo camera is used will be described here.

FIG. 6 is a diagram showing an overall picture of a pig farming environment in which the pig breeding support device according to the second example of the present embodiment is adopted. The configuration shown in FIG. 6 differs from the configuration shown in FIG. 1 in that the camera unit 211 employs a stereo camera and that the angle of view Fr, which is the imaging range, is wider. Since the camera unit 211 has a wide angle of view Fr, it can image many pigs 302 at once. Although FIG. 6 shows the angle of view Fr only for the upper left pen 301, the camera unit 211 that overlooks the other pens 301 can also image each pen 301 at the same angle of view.

FIG. 7 is a diagram illustrating a method of generating a buttock image when the camera unit 211 is employed. Many pigs 302 are often reflected in the captured image captured by the camera unit 211. Therefore, the detection unit 112 cuts out the entire buttock region of the pig 302 whose buttocks are directly facing each other from one side of the stereo image. Specifically, similarly to the method described using FIG. 3, an image area where the tail is located at the center of the buttock in the width direction is cut out and used as the buttock image. At this time, the image size of the cutout image is not a fixed size, but is adjusted individually so that the left and right body boundaries of each pig are included.

For example, when three buttock images Pic1 to Pic3 are extracted as shown in the figure, the detection unit cuts out the corresponding regions from the other stereo image. Then, distances d1, d2, and d3 from the camera unit 211 to the respective buttocks are calculated using the amount of deviation between the two corresponding buttock images. The detection unit 112 calculates the hip width from one of the stereo images as described using FIG. 3, and uses the respective distances to correct the hip width for the standard distance d ₀ (for example, 1 m). By making corrections in this way, it is possible to uniformly compare and verify the width of the buttocks. The processing after detecting the hip width is the same as in the first embodiment.

Next, a pig breeding support device according to a third example of this embodiment will be described. The pig breeding support devices of the first and second embodiments detected pigs with growth retardation by observing the width of the rump, which has a strong correlation with the pig's weight. The pig breeding support device according to the third embodiment detects pigs with growth retardation by observing the pig's abdomen. In the following description, components that perform the same functions as those in the first and second embodiments will be given the same reference numerals, and the description thereof will be omitted unless specifically mentioned.

FIG. 8 is a diagram showing an overall picture of a pig farming environment in which the pig breeding support device according to the third example of the present embodiment is adopted. The camera unit 212 is installed for each pen 301, and is installed so as to be suspended from the ceiling, for example, so as to be able to take an overview of the entire pen 301 that is the object of observation. The camera unit 212 converts the captured image into image data and transmits it to the server 100' via the network 200.

FIG. 9 is a diagram showing the hardware configuration of a server 100' and peripheral devices as a pig breeding support device. The detection unit 112' extracts the pig's abdomen from the image of the image data received from the acquisition unit 111, and determines the width and narrowness, which is the ratio of the widest abdominal width to the narrowest abdominal width, based on the extracted abdominal image. Detect the ratio. Note that such a wide/narrow ratio may hereinafter be referred to as an abdominal wide/narrow ratio.

When the detection neural network 122 used by the detection unit 112' receives the image captured by the camera unit 210, it searches for an image area of a pig that shows a suitable figure for detecting the abdominal wideness/narrowness ratio, and searches for an image area in the image area. Outputs the abdominal wide/narrow ratio of the pig in the photo. Note that in the first and second embodiments, the calculation unit 113 and the notification unit 114 calculated statistical values and issued notifications for the buttock width, but in the third embodiment, the abdominal wide/narrow ratio Calculate statistical values and issue notifications.

Next, the abdominal wide/narrow ratio will be explained. As a result of further research on a method to more easily evaluate the growth rate of pigs housed in pens, the inventor of the present application found that there is a strong correlation between the abdominal width ratio of pigs and the weight of the pigs. Ta. Similar to hip width, the correlation between abdominal wideness and narrowness ratio and body weight was found to have little variation between individuals, and was particularly stronger than the correlations between other body parts and body weight. It was also found that observation of the pig's abdomen is easy, and analysis using images can be performed accurately and easily. Therefore, in this embodiment, by observing the abdomens of pigs that move around randomly within the pen and detecting the abdominal wide/narrow ratio, it is determined whether or not there are pigs with growth retardation in the breeding group.

FIG. 10 is a diagram illustrating the difference in abdominal wide/narrow ratio between normal pigs and lean pigs. The abdominal wide-narrow ratio is the length along the gravity direction from the upper end of the back to the lower end of the abdomen near the hind legs (or front legs) when the pig to be observed is standing on all four legs. It refers to b/a, where b is the length along the gravity direction from the upper end of the back to the lower end of the abdomen where it hangs down the most. For example, b/a of the normal pigs shown in FIG. 10(A), which are many pigs that are growing normally in the same pen 301, is the pig with delayed growth shown in FIG. 10(B). greater than b/a of lean pigs. Therefore, by performing the same statistical processing as the above-mentioned buttock width using this abdominal wide/narrow ratio as a parameter, it can be determined whether the pig in the acquired image is a pig with growth retardation.

FIG. 11 is a diagram illustrating a procedure for detecting the abdominal wide/narrow ratio using the detection neural network 122, which is a learning model. When the detection unit 112 ′ receives an acquired image that is an overhead image of the pen 301 from the acquisition unit 111 , the detection unit 112 ′ inputs the acquired image to the detection neural network 122 .

The detection neural network 122 is a teacher that learns by giving a considerable amount of teacher data that is associated with the bird's-eye view of the pen 301, with a rectangular area surrounding a pig standing on four legs as the correct answer. It has been created in advance through ``arithmetic learning''. The training data is created, for example, by an operator extracting a pig standing on all four legs from each pig in the bird's-eye view image, surrounding it with a rectangular area, and giving the correct answer. If a plurality of target pigs exist in one bird's-eye view image, the operator surrounds each of them with a rectangular area and designates them as correct. The detection neural network 122, which is a learning model generated in this way, is incorporated into the server 100, which is a pig breeding support device, and is made available for use.

For example, if a pig farm is provided with eight pens 301 to be observed, image data is sequentially sent to the server 100 from eight camera units 212 that overlook each pen 301. When the acquisition unit 111 acquires these images, they are delivered to the detection unit 112'. The detection unit 112' sequentially inputs the images read out from the storage unit 120 to the detection neural network 122. The detection neural network 122 executes arithmetic processing each time an image is input, and extracts from the image a region in which a pig standing on four legs is shown as a verification area. In this way, for example, the verification areas "area01", "area02", etc. are extracted from the input image as shown in the figure.

The detection unit 112' calculates the abdominal wide/narrow ratio of the pig in each of the extracted verification areas. Specifically, similar to the calculation of the hip width explained using Fig. 4, the narrowest width is obtained by counting the number of pixels from the body boundary on the back side to the body boundary on the ventral side with the base of the hind legs as a reference. Determine the abdominal width a. Furthermore, the straight line of the abdomen width a is gradually shifted parallel to the front leg side, and each time the number of pixels from the body boundary on the back side to the body boundary on the abdominal side is counted. Then, the largest one among them is determined as the widest abdominal width b. After determining a and b in this manner, the abdominal wide/narrow ratio b/a is calculated. If a plurality of verification areas are extracted, this calculation is repeated for each verification area, and the obtained abdominal wide/narrow ratio b/a is delivered to the calculation unit 113 and the notification unit 114.

When a predetermined number of abdominal wide/narrow ratios are detected in the pen 301 to be observed, the statistical value calculated by the calculation unit 113 can be used to represent the characteristics of the entire pig kept in the pen 301. The prescribed number is set to, for example, 80% of the number of pigs housed in the pen 301. If a value of around 80% is set, it can be expected that even if a single pig is observed multiple times, the effect will be kept small. If the subsequently detected abdominal wide/narrow ratio shows a deviation of more than a threshold value from the calculated statistical value, it can be assumed that the pig for which the abdominal wide/narrow ratio was detected has a growth abnormality. . The method of estimating growth abnormalities through statistical processing is the same whether using the hip width or the abdominal wide/narrow ratio.

Next, the processing procedure of the pig rearing support method using the server 100' will be explained. FIG. 12 is a flow diagram illustrating the processing procedure of the calculation unit 110. The flow starts from the moment an instruction to start observation is received.

The acquisition unit 111 acquires image data of img _n from the camera unit 210 of the n-th pen via the communication unit 130 in step S201. For example, if n=1, image data of img ₁ is acquired from the camera unit 210 directed toward the first pen. The acquisition unit 111 delivers the acquired image data to the detection unit 112'. n is incremented each time step S201 is executed.

In step S202, the detection unit 112' inputs the image img _n of the received image data to the detection neural network 122, as described using FIG. If the verification area is extracted from the output, the process advances to step S203, and if it is not extracted, the process returns to step S201.

After proceeding to step S203, the detection unit 112' calculates the torso wide/narrow ratio for each extracted verification area. In other words, the torso wide/narrow ratio is detected from the partial image extracted as the verification area. The detection unit 112' delivers the result to the calculation unit 113 and notification unit 114. When the calculation unit 113 receives the newly detected trunk wide/narrow ratio from the detection unit 112', it updates the statistical value calculated from the previously received trunk wide/narrow ratio. The calculation unit 113 then proceeds to step S205 and checks whether the number of abdominal wide/narrow ratios received from the time point when observation was started exceeds a specified number. If the number exceeds the specified number, the statistical value calculated in step S204 and information indicating that the specified number has been exceeded are transferred to the notification unit 114, and the process proceeds to step S206. If the number has not been exceeded, the process returns to step S201.

Proceeding to step S206, the notification unit 114 determines that the abdominal wide/narrow ratio received from the detection unit 112' is greater than or equal to the preset standard deviation from the statistical value received from the calculation unit 113. Check whether it is a value. In this embodiment, it is checked whether the abdominal wide/narrow ratio received from the detection unit 112' is less than the average value by α×σ, and if it is, the process proceeds to step S207; if not, the process proceeds to step S207. Skip S207 and proceed to step S208. If the process proceeds to step S207, the notification unit 114 sends a notification to the keeper terminal 220 that there is a pig that is considered to be developmentally delayed, with the pen information of the detected pen added. The pen information is, for example, a pen number given to identify a pen. Once the notification is issued, the process advances to step S208. In step S208, the calculation unit 110 checks whether the observation period has ended. If the observation period has not ended, the process returns to step S201 and continues observation. If it has been completed, the series of processing is completed.

Note that in the above processing procedure, similarly to the processing procedure for the buttock width, each of the abdominal wide/narrow ratios accumulated up to that time may be determined after the predetermined number has been exceeded. Further, the process for setting the observation period and starting a new observation period is also similar.

In this embodiment described through several examples above, it is assumed that a plurality of pens 301 are provided in one pigsty, but it is also possible to observe pens 301 provided across a plurality of pigsty. It's okay. Further, the output destination of the detection notification is not limited to the zookeeper terminal 220. For example, one notification light connected to the servers 100, 100' via the wireless unit 230 and the network 200 may be installed adjacent to each pen 301, and notifications may be output to the notification light. For example, when the server 100, 100' detects a pig with growth retardation in the fifth pen, the server 100, 100' sends a notification to a notification light installed adjacent to the fifth pen and turns on the notification light. Further, the calculation unit 110 sends the results detected by the detection units 112 and 112′, the statistical values calculated by the calculation unit 113, and the notification issued by the notification unit 114 to the display monitor 150 connected to the servers 100 and 100′. It may be displayed directly.

Further, in the present embodiment described above, a case has been described in which the servers 100 and 100' function as a pig breeding support device, but the hardware configuration is not limited to this. If the mobile terminal described as the breeder terminal 220 performs the same processing as the servers 100 and 100', the mobile terminal can function as a pig breeding support device. Furthermore, for example, if a part of the processing of the servers 100, 100' is configured to be carried out by the breeder terminal 220, a system in which the servers 100, 100' and the breeder terminal 220 cooperate can become a pig breeding support device.

100, 100'...Server, 110...Computation unit, 111...Acquisition unit, 112, 112'...Detection unit, 113...Calculation unit, 114...Notification unit, 120...Storage unit, 121, 122...Neural network for detection, 130 ...Communication unit, 150...Display monitor, 160...Input device, 200...Network, 210, 211, 212...Camera unit, 220...Keeper terminal, 230...Wireless unit, 301...Pen, 302...Pig, 303...Feeding container , 304...Water supply container

Claims (9)

an acquisition unit that acquires image data of an image captured by a camera installed toward a pen in which pigs are reared in groups;
a detection unit that extracts the buttock of the pig from the image of the image data and detects the width of the buttock of the pig based on the extracted buttock image;
a calculation unit that calculates a statistical value of the buttock width detected by the detection unit during a set period;
A pig breeding support device comprising: a notification section that issues a notification when the detection section detects the rump width that shows a deviation of more than a threshold value from the statistical value. The pig breeding support device according to claim 1, wherein the detection unit extracts the rump using a learning model learned from a teacher image showing a pig with its tail exposed. an acquisition unit that acquires image data of an image captured by a camera installed toward a pen in which pigs are reared in groups;
a detection unit that extracts the abdomen of the pig from the image of the image data and detects a width ratio that is a ratio between the widest abdominal width and the narrowest abdominal width based on the extracted abdominal image;
a calculation unit that calculates a statistical value of the wide/narrow ratio detected by the detection unit during a set period;
A pig breeding support device comprising: a notification unit that issues a notification when the detection unit detects the wide/narrow ratio that shows a deviation of a threshold value or more from the statistical value. The pig breeding support device according to claim 3, wherein the detection unit extracts the abdomen using a learning model learned from a teacher image showing a pig standing on four legs. The pig breeding support device according to any one of claims 1 to 4, wherein the notification unit also notifies pen information regarding the detected pen when a plurality of the pens are provided. an acquisition step of acquiring image data of an image taken by a camera installed toward a pen in which pigs are reared in groups;
a detection step of extracting the buttock of the pig from the image of the image data and detecting the width of the buttock of the pig based on the extracted buttock image;
a calculation unit that calculates a statistical value of the buttock width detected by the detection step during a set period;
A notification step of issuing a notification when the detection step detects the rump width that shows a deviation of more than a threshold value from the statistical value. an acquisition step of acquiring image data of an image taken by a camera installed toward a pen in which pigs are reared in groups;
a detection step of extracting the abdomen of the pig from the image of the image data and detecting a wide-narrow ratio that is a ratio of the widest abdominal width to the narrowest abdominal width based on the extracted abdominal image;
a calculation unit that calculates a statistical value of the wide/narrow ratio detected by the detection step during a set period;
A notification step of issuing a notification when the wide/narrow ratio exhibiting a deviation of more than a threshold value from the statistical value is detected in the detection step. an acquisition step of acquiring image data of an image taken by a camera installed toward a pen in which pigs are reared in groups;
a detection step of extracting the buttock of the pig from the image of the image data and detecting the width of the buttock of the pig based on the extracted buttock image;
a calculation unit that calculates a statistical value of the buttock width detected by the detection step during a set period;
A pig breeding support program that causes a computer to execute a notification step of issuing a notification when the rump width exhibiting a deviation of a threshold value or more from the statistical value is detected in the detection step. an acquisition step of acquiring image data of an image taken by a camera installed toward a pen in which pigs are reared in groups;
a detection step of extracting the abdomen of the pig from the image of the image data and detecting a wide-narrow ratio that is a ratio of the widest abdominal width to the narrowest abdominal width based on the extracted abdominal image;
a calculation unit that calculates a statistical value of the wide/narrow ratio detected by the detection step during a set period;
A pig breeding support program that causes a computer to execute a notification step of issuing a notification when the wide/narrow ratio showing a deviation of a threshold value or more from the statistical value is detected in the detection step.

Images