[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

WO2022153829A1 - Pig rearing assistance apparatus, pig rearing assistance method, and pig rearing assistance program - Google Patents

Pig rearing assistance apparatus, pig rearing assistance method, and pig rearing assistance program Download PDF

Info

Publication number
WO2022153829A1
WO2022153829A1 PCT/JP2021/047913 JP2021047913W WO2022153829A1 WO 2022153829 A1 WO2022153829 A1 WO 2022153829A1 JP 2021047913 W JP2021047913 W JP 2021047913W WO 2022153829 A1 WO2022153829 A1 WO 2022153829A1
Authority
WO
WIPO (PCT)
Prior art keywords
pig
posture
specific posture
pen
counting
Prior art date
Application number
PCT/JP2021/047913
Other languages
French (fr)
Japanese (ja)
Inventor
翔 伊澤
直樹 森下
慎 助川
大介 内田
雅貴 奥田
寛樹 本橋
翔 本間
尚昇 東尾
Original Assignee
日本ハム株式会社
株式会社エヌ・ティ・ティ・データ
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本ハム株式会社, 株式会社エヌ・ティ・ティ・データ filed Critical 日本ハム株式会社
Priority to JP2022575504A priority Critical patent/JPWO2022153829A1/ja
Publication of WO2022153829A1 publication Critical patent/WO2022153829A1/en

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
    • A01K—ANIMAL HUSBANDRY; AVICULTURE; APICULTURE; PISCICULTURE; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
    • A01K29/00—Other apparatus for animal husbandry
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the present invention relates to a pig breeding support device, a pig breeding support method, and a pig breeding support program.
  • a system for detecting abnormalities in livestock is known. For example, when various sensors are attached to individual livestock and the output is a value that cannot be in a healthy state, the individual is judged to be abnormal (see, for example, Patent Document 1).
  • the present invention has been made to solve such a problem, and it is possible to tell the keeper in a timely manner which pen has a sick pig without undue cost and labor. It provides a pig breeding support device and the like that can inform.
  • the pig breeding support device includes an acquisition unit that acquires image data of an image captured by a camera installed toward a plurality of pens in which pigs are group-reared, and an image. It includes a detection unit that detects a specific posture of a pig based on an image of data, and a counting unit that counts the number of specific postures detected by each of a plurality of pens during a set observation time.
  • the pig breeding support method includes an acquisition step of acquiring image data of an image captured by a camera installed toward a plurality of pens in which pigs are group-reared in each. It has a detection step of detecting a specific posture of a pig based on an image of image data, and a counting step of counting the number of times of the specific posture detected by each of a plurality of pens during a set observation time.
  • the pig breeding support program includes an acquisition step of acquiring image data of an image captured by a camera installed toward a plurality of pens in which pigs are group-reared in each.
  • a detection step that detects a specific posture of a pig based on an image of image data and a counting step that counts the number of specific postures detected by each of a plurality of pens during a set observation time are executed on a computer. Let me.
  • the present invention it is possible to determine which of the plurality of pens in which pigs are bred in a group has a pig that is in poor physical condition by utilizing the habit of the pig, thereby causing excessive cost and labor. It is possible to provide a pig breeding support device or the like that can notify the keeper in a timely manner without needing it.
  • FIG. 1 is a diagram showing an overall picture of a pig farming environment that employs the pig breeding support device according to the present embodiment.
  • the pig farm is equipped with a plurality of pens 301 separated by walls and fences.
  • Each pen 301 accommodates a plurality of pigs 302 (for example, about 10 pigs) and is bred in a group.
  • the number of pigs 302 raised by each pen 301 can be adjusted according to the breed of pig 302, the breeding environment, and the like.
  • a camera unit 210 for observing the housed pig 302 is installed for each pen 301.
  • the camera unit 210 is installed so as to be suspended from the ceiling, for example, toward the pen 301 so that the entire pen 301 to be observed can be viewed from a bird's-eye view.
  • the camera unit 210 converts the captured image into image data and transmits the captured image to the server 100 via the network 200.
  • the wireless unit 230 installed in 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.
  • it may be configured to establish communication by a wired connection.
  • the network 200 connecting the camera unit 210 and the server 100 may use the Internet or an intranet, and when the management facility where the server 100 is installed is provided in the pig farm, short-range wireless communication is performed. It may be adopted.
  • the keeper who takes care of the pig 302 can carry the keeper terminal 220.
  • the keeper terminal 220 is, for example, a tablet terminal or a smartphone, and can exchange various information with and from the server 100 via the wireless unit 230 and the network 200.
  • the keeper can input the breeding record to the keeper terminal 220 and transfer it to the server 100, or can call up the breeding record stored in the server 100.
  • the keeper terminal 220 receives the excess notification described later, it displays that fact.
  • a server 100 as a pig breeding support device is installed in the management facility.
  • the server 100 is connected to the network 200.
  • the server 100 sequentially acquires image data from the camera unit 210 installed for each pen 301, detects a specific posture of the pig 302 based on the image data, and counts the number of times. The specific posture will be described later.
  • the server 100 can display the result on the display monitor 150 in response to a request from the administrator or the keeper.
  • the display monitor 150 is, for example, a monitor including a liquid crystal panel.
  • the server 100 is connected to an input device 160 that accepts operations by an administrator or a keeper.
  • the input device 160 is composed of a keyboard, a mouse, a touch panel superimposed on the display surface of the display monitor 150, and the like.
  • the pig breeding support device in the present embodiment detects at least one pig that is likely to be in poor physical condition by detecting a specific posture that the pig 302 in the pen 301 tends to show when it is in poor physical condition. Support the breeding work by notifying the keeper of the pen.
  • the keeper When there are a plurality of pens 301 in the pig farm as shown in FIG. 1, the keeper is informed by the pig breeding support device of information on which pen 301 contains the pig 302 showing a sign of poor physical condition. You can save a lot of effort just by doing this. That is, the keeper does not have to perform the extraction work of the pig 302 showing a sign of poor physical condition until notified by the pig breeding support device, and even when notified, the keeper may perform the extraction work for the specific pen. If the presence of pigs 302, which show signs of poor health for a limited number of pigs in a pen, is shown, even inexperienced breeders are relatively easy to extract.
  • the pig breeding support device in the present embodiment detects a specific posture of the pig 302 by using an image captured by the camera unit 210.
  • the specific posture is the posture that the pig tends to show when he / she is in poor physical condition, and the applicant has obtained some of the specific postures as knowledge.
  • FIG. 2 is a diagram illustrating an example of a specific posture.
  • FIG. 2A shows a dog-sitting posture, specifically, a sitting posture with the front legs upright.
  • the hind legs are often bent and in contact with the abdomen, and the head may be lifted or hung down.
  • the dog-sitting posture is one of the postures that pigs tend to show when they are in poor physical condition, and is treated as a specific posture in this embodiment.
  • FIG. 2B shows a dozing posture, specifically, a posture in which the front legs are placed under the prone torso. That is, the posture is such that the front legs are folded and sandwiched between the floor surface and its own torso, and the head is often in contact with the floor surface and is quiet.
  • the eaves posture is one of the postures that the pig tends to show when he / she is in poor physical condition, and is treated as a specific posture in this embodiment.
  • Fig. 2 (C) shows the back bay posture, specifically, the posture in which the back is curled up and leans forward. At this time, the distance between the front and rear leg tips is narrower than the distance between the leg bases.
  • the back bay posture is one of the postures that pigs tend to show when they are in poor physical condition, and is treated as a specific posture in this embodiment.
  • FIG. 2 (D) is punting, specifically, a posture accompanied by movement, in which the abdomen is wavy in the lying position.
  • the lying position itself is a posture that is also seen in healthy pigs, but unwell pigs may show the movement of repeatedly raising and lowering the abdomen and hitting it against the floor. Punting is one of the movement-associated postures that pigs tend to show when they are in poor physical condition, and is treated as a specific posture in this embodiment.
  • the pig breeding support device determines the specific posture of the pig 302 group-reared based on the image captured by the camera unit 210 installed toward the pen 301. Detect. Each camera unit 210 is installed so as to look down in the inclined direction so that the posture of the pig 302 housed in the pen 301 to be imaged can be captured.
  • a specific posture can be detected by, for example, extracting the contour line of each pen from the image overlooking the pen 301 and pattern matching the shape of each contour line with the template pattern.
  • the template pattern is a contour line that captures each posture from various angles, and is prepared in advance.
  • the dog-sitting posture, the eaves posture, and the back bay posture which can be detected even if there is no change in movement, can be detected from a single image of a still image, but the punting, which is a movement-accompanying posture, is front and back. Detect using multiple still images and moving images.
  • FIG. 3 is a diagram illustrating a procedure for detecting a specific posture using the detection neural network 121 which is a learning model.
  • the detection neural network 121 associates the image of the pig in the dog-sitting posture with the teacher data as the correct answer, and the image of the pig in the dozing posture with the "sleeping posture" as the correct answer.
  • the input image of the detection neural network 121 may be a still image as long as the posture accompanied by the change in movement is not added as the specific posture.
  • the detection neural network 121 which is the learning model generated in this way, is incorporated into the server 100, which is a pig breeding support device, and is used. Specifically, for example, assuming that eight pens 301 to be observed are provided in the pig farm, image data is sequentially sent to the server 100 from the eight camera units 210 that overlook each pen 301. The images img 1 to img 8 of the image data are sequentially input to the detection neural network 121.
  • the detection neural network 121 outputs the number of pigs determined to have a specific posture in the image as the number of detections. Specifically, the number of dogs in the dog-sitting posture, the number of heads in the eaves posture, the number of heads in the dorsal bay posture, and the number of heads in the punting posture are output as the detected number of each posture in the image.
  • the number of detections for the image img 1 of the first pen is the dog sitting posture "0", the dozing posture "1", the back bay posture "1", the punting "1", and the image of the second pen.
  • the number of detections for img 2 the number of detections of each posture for each pen 301 is obtained, such as the dog sitting posture "2", the dozing posture "0", the back bay posture “0", and the punting "0". ..
  • the server 100 repeats such a process for a preset observation time, and counts and integrates the number of detections for each pen 301.
  • FIG. 4 is a diagram showing the hardware configuration of the server 100 as a pig breeding support device and peripheral devices.
  • the server 100 can be connected to the display monitor 150, the input device 160, the camera unit 210, and the keeper terminal 220.
  • the server 100 is mainly composed of a calculation unit 110, a storage unit 120, and a communication unit 130.
  • the arithmetic unit 110 is a processor (CPU: Central Processing Unit) that controls the server 100 and executes a program.
  • 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 breeding support program stored in the storage unit 120 and executes various processes related to the pig breeding support.
  • the storage unit 120 is a non-volatile storage medium, and is composed of, for example, an HDD (Hard Disk Drive).
  • the storage unit 120 can store various parameter values, functions, display element data, lookup tables, and the like used for control and calculation, in addition to a program that executes control and processing of the server 100.
  • the storage unit 120 stores, in particular, the detection neural network 121 and the counting list 122. As described above, when the image captured by the camera unit 210 is input, the detection neural network 121 outputs the number of detections representing the number of heads of a specific posture existing in the image.
  • the counting list 122 is a record relating to counting of a specific posture, and will be described in detail later.
  • the storage unit 120 may be composed of a plurality of hardware.
  • the storage medium for storing the program and the storage medium for storing the detection neural network 121 may be composed of different hardware.
  • the communication unit 130 includes, for example, a LAN unit, transmits an imaging control signal generated by the calculation unit 110 to the camera unit 210, and transmits image data sent from the camera unit 210 to the calculation unit 110 via the network 200. I will hand it over.
  • the transfer of data executed between the keeper terminal 220 and the calculation unit 110 is relayed.
  • the communication unit 130 can also relay the exchange of data and control signals with other external devices. For example, it can also be used when fetching update data of a pig breeding support program or a 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, and a counting unit 113.
  • the acquisition unit 111 mainly acquires the image data of the image captured by the camera unit 210 and delivers it to the detection unit 112.
  • the detection unit 112 mainly detects the specific posture of the pig 302 based on the image of the image data received from the acquisition unit 111, and delivers the result to the counting unit 113.
  • the counting unit 113 mainly counts the number of times of the specific posture detected by each of the plurality of pens 301 during the set observation time.
  • FIG. 5 is a diagram for explaining a reference observation time as a target period for the counting unit 113 to count the number of times of a specific posture.
  • the counting unit 113 resets the number of times counted for each preset reference observation time.
  • the reference observation time is set to 24 hours.
  • the counting unit 113 counts how many times a specific posture is detected in each pen 301 during this reference observation time.
  • the detection unit 112 uses the detection neural network 121 to use the images img 1 to img 8 of the image data of each pen 301 periodically acquired by the acquisition unit 111.
  • the number of heads (detected number) that take a specific posture in each image is detected. This is repeated during the reference observation time, and the number of times a specific posture is detected by each pen 301 is counted.
  • a keeper approach period is set as such an exclusion period.
  • the keeper approach period is set as a fixed period including a scheduled time from when the keeper approaches the pen 301 to when the keeper approaches the pen 301.
  • the keeper approach period is set from 8:00 to 9:00 when the keeper enters the pen 301 for feeding, for example.
  • a period during which the keeper approaches for cleaning the pen 301, inspecting the pig 302, or the like may also be set as the keeper approach period.
  • the exclusion period may be set by adding a period before and after the event implementation period, a period in which the pig 302 senses the event and begins to get excited, and a period in which the pig 302 regains calm after the event.
  • the exclusion period can be set in advance by the administrator or the like by operating the input device 160.
  • the exclusion period does not have to be included in each reference observation time. For example, if it is an event in which a boar is put into a pen for raising a sow for a short period of time, the reference observation on a specific day when the event is held It may be set only for time.
  • the counting unit 113 excludes the exclusion period set in this way from the target of counting. Specifically, processing such as stopping counting by the counting unit 113, stopping the detection of a specific posture by the detection unit 112, stopping the acquisition of image data by the acquisition unit 111, and stopping the imaging by the camera unit 210 is performed. Can be adopted.
  • one specific posture is detected as one time in consideration of the standard duration of the pig 302 continuing the specific posture. Adjusted and set to.
  • the breed of pig 302 the breeding environment, the age of the moon, and the like can be taken into consideration when setting the cycle.
  • the counting unit 113 manages the number of times of the specific posture counted for each pen 301 by the counting list 122.
  • FIG. 6 is a diagram illustrating an example of the counting list 122.
  • One counting list 122 is generated at one reference observation time, and is updated as appropriate during the observation time.
  • the counting list 122 includes the notification threshold value and the observation date.
  • the notification threshold value is a value preset by the administrator or the like, and when the total number of times of the specific posture counted by each pen 301 exceeds this notification threshold value, the counting unit 113 informs the keeper terminal 220 or the like to that effect. Output an excess notification to notify.
  • the manager or the like sets the notification threshold value in consideration of the breed and breeding environment of the pig 302, particularly the number of pigs 302 housed in one pen 301. More specifically, a threshold value at which it can be determined that at least one pig 302 showing a sign of poor physical condition exists in the target pen 301 is set based on the statistics and experience so far. In the example of FIG. 6, "50 times" is set.
  • the observation date is the date on which the observation is performed, and when the counting list 122 is referred to after the observation, it represents the date on which the observation was performed. If the reference observation time is less than 24 hours, for example, the observation time at which the observation was started may be added.
  • the counting list 122 includes a tabulation table showing the number of times of a specific posture counted for each pen 301.
  • the tabulation table shows the number of times the pen numbers (for example, 8 from the 1st pen to the 8th pen), the number of times each of the specific postures (dog sitting posture, eaves posture, back bay posture, punting) were counted, and those. It consists of the total number of times and flag information.
  • the counting unit 113 confirms that the pen 301 whose total number of times counted exceeds the notification threshold has appeared, it generates an excess notification and outputs it to the keeper terminal 220 or the like.
  • the counting unit 113 adds pen information regarding the specific pen whose number of times counted exceeds the notification threshold value to the excess notification and outputs the pen information.
  • the pen number is added as pen information.
  • "notified" information is recorded as flag information.
  • FIG. 7 is a diagram showing a display example of the keeper terminal 220 that has received the excess notification.
  • the keeper terminal 220 receives the excess notification and displays the content on the display panel. Specifically, as shown in the figure, the keeper terminal 220 refers to the pen information added to the excess notification and displays the pen number in which the number of times of the specific posture exceeds the notification threshold (specified value) (FIG. In the example of, "second pen"). In addition, if an indoor map regarding the pen arrangement of the pig farm is held, it is also displayed so that the position of the specific pen is recognized. The keeper terminal 220 may make such a display and emit a notification sound.
  • FIG. 8 is a flow chart illustrating a processing procedure of the calculation unit 110. The flow starts from the start of the reference observation time. Here, the processing for the exclusion period will be omitted.
  • the counting unit 113 starts aging by the aging timer T in step S101 as an initial process, and resets all the counters Cn for counting the total number of times of the specific posture of the pig 302 housed in the nth pen. ..
  • the number of pens in the pig farm is m, and counters from C 1 to C m corresponding to each pen are prepared.
  • the counting unit 113 proceeds to step S102 and sets the variable n to 1.
  • the counter to be processed is switched to C1 accordingly.
  • the acquisition unit 111 acquires image data of img n from the camera unit 210 directed to the nth pen via the communication unit 130.
  • the acquisition unit 111 delivers the acquired image data to the detection unit 112.
  • step S104 the detection unit 112 inputs the image im g n of the received image data into the detection neural network 121 read from the storage unit 120, and the number of pigs 302 (detection number) taking a specific posture in the image im g n . ) Is output. Specifically, as described with reference to FIG. 3, the number of dogs in the dog-sitting posture, the number of heads in the eaves posture, the number of heads in the back bay posture, and the number of wheezing heads are output. The detection unit 112 delivers the number of these heads to the counting unit 113 as the number of detections.
  • step S105 the counting unit 113 reads out the counting list 122, adds each detected number received according to each posture, and adds the detected number received from the detection unit 112 to the counter value of Cn at that time.
  • Cn is updated by adding all the numbers of.
  • step S106 the counting unit 113 determines whether or not the updated value of Cn exceeds the notification threshold value Cd . If it is determined that the notification threshold value C d has been exceeded, the process proceeds to step S107 to generate an excess notification and output it to the keeper terminal 220. At this time, it is added that the nth pen is a specific pen as pen information. After outputting the excess notification, the process proceeds to step S108. If the counting unit 113 determines in step S106 that the notification threshold value C d has not been exceeded, the counting unit 113 skips step S107 and proceeds to step S108.
  • the counting unit 113 increments the variable n in step S108, and proceeds to step S109.
  • step S109 it is determined whether or not the variable n exceeds the number of pens m in the pig farm. If it is determined that the value is not exceeded, the process returns to step S103, and the same processing is executed for the incremented variable n. If it is determined that the value has been exceeded, the process proceeds to step S110.
  • the counting unit 113 determines whether or not the time-lapse timer T has exceeded the reference observation time T c . If it is determined that the value is not exceeded, the process returns to step S102 at intervals corresponding to the preset period. When it is determined that the amount has been exceeded, a series of processes are terminated. When the observation is continuously executed, the process is started again from step S101.
  • FIG. 9 is a diagram illustrating a counting list 122'of a pig breeding support device according to another embodiment.
  • the counting list 122' is different from the counting list 122 shown in FIG. 6 in that it has a weighting coefficient table.
  • the dog sitting posture, the eaves posture, the back bay posture, and the punting which are the targets of detection as specific postures, are treated equally, and the number of times detected in each posture is simply added up. The total number of times was counted.
  • the degree to which each posture is detected to indicate a sign of poor physical condition in other words, the degree of contribution to the judgment of poor physical condition may differ from each other.
  • weighting is applied according to each posture detected as a specific posture, and the total number of times is counted.
  • the weighting coefficient of the dog sitting posture is "1.0"
  • the weighting coefficient of the dozing posture is "1.0”
  • the weighting coefficient of the dorsal bay posture is "0". It is "0.8” and the weighting coefficient of punting is "1.6".
  • the contribution of the back bay posture to the judgment of poor physical condition is lowered, and the contribution of punting is increased.
  • the manager or the like can set the weighting coefficient in advance according to the degree of contribution in each posture that is empirically and statistically grasped.
  • the total number of times of the second pen observed that the dog sitting posture is 4 times
  • the eaves posture is 3 times
  • the back bay posture is 4 times
  • the punting is 1 time
  • weighting is given according to each posture to be detected, but a method of giving weight to the duration in which each posture is detected can also be adopted. ..
  • the number of times of counting is different depending on whether the posture is solved in 1 minute or continued for 10 minutes.
  • the detection time (minutes) x 1.0 is set for 1 minute from the start of detection
  • the detection time (minutes) x 1.2 is set for up to 3 minutes thereafter
  • weighting by posture may also be applied.
  • FIG. 10 is a diagram showing an overall picture of a pig farming environment in which a pig breeding support device according to another embodiment is adopted.
  • the same elements as those in FIG. 1 are given the same numbering, and the description thereof will be omitted.
  • the keeper does not have the keeper terminal 220, and instead, one notification light 240 is installed adjacent to each pen 301.
  • Each notification light 240 is connected to the server 100 via the wireless unit 230 and the network 200.
  • the server 100 transmits a notification signal corresponding to the excess notification to the notification light 240 installed adjacent to the fifth pen, and the notification light. Turn on 240.
  • the keeper can recognize the pen 301 to be directed even if he / she does not have the keeper terminal 220.
  • the keeper may search for a pig showing signs of poor physical condition by targeting the pig 302 housed in the pen 301 in which the notification light 240 is turned on.
  • FIG. 11 is a diagram for explaining the counting list 122 ”of the pig breeding support device according to still another embodiment.
  • the plurality of pigs 302 housed in one pen 301 are Therefore, if any of the pigs 302 housed in the specific pen 301 to be observed takes a specific posture, it is detected as one specific posture.
  • the notification threshold When is set to 50 times, the counting unit regardless of whether one pig 302 takes a specific posture exceeding 50 times or 10 pigs 302 take a specific posture 5 to 6 times each. 113 outputs an excess notification. That is, even if the keeper knows which pen 301 has detected a specific posture exceeding the specified value, is the specific pig 302 strongly showing a sign of poor physical condition? It is necessary to determine whether the pig 302 in the pen tends to be in poor physical condition as a whole.
  • each pen 301 a technique for distinguishing and recognizing a plurality of pigs 302 housed in each pen 301 has become known.
  • individual identification can be realized by attaching an identification marker to each pig 302 and analyzing an image obtained by capturing the image with the camera unit 210.
  • each pig 302 is imaged and an identification number is associated with the captured image, and then the pig 302 that detects a specific posture is associated with any identification number. May be detected using a learning model.
  • the counting list 122 ”shown in FIG. 11 is a counting list when the individual pigs 302 housed in each pen 301 can be identified. For example, 10 pigs 302 are housed in each pen 301. Each pig 302 is distinguished by an assigned identification number. Then, the detection unit 112 detects a specific posture and identifies the identification number of the pig 302 that has taken the specific posture. The 113 receives the information from the detection unit 112 and updates the number of times of the specific posture of the pig corresponding to the specified identification number.
  • the notification threshold value is set to a smaller value when individual identification is performed than when individual identification is not performed, it is possible to quickly detect poor physical condition of a specific pig 302.
  • the notification threshold is set to 25 times.
  • the keeper terminal sends an excess notification to which the information of the identification number is added. Output to 220.
  • the notification light 240 described with reference to FIG. 10 includes a display unit, the counting unit 113 outputs an excess notification to the notification light 240 adjacent to the pen in which the specific pig 302 is housed. Then, the identification number may be displayed on the display unit of the notification light 240.
  • the keeper can also obtain information on the specific pig 302, the keeper may select the pig 302 from among the plurality of pigs 302 housed in the pen 301. A specific pig 302 can be easily found.
  • one camera unit 210 is installed for each pen 301, but a camera unit for overlooking a plurality of pens 301 may be installed.
  • the acquisition unit 111 may divide the image acquired from the camera unit along the boundary of each pen 301, and sequentially deliver each of the divided images to the detection unit 112.
  • a plurality of camera units 210 may be installed for one pen 301. For example, if a camera unit 210 that captures the pig 302 in the pen 301 from the side is installed, erroneous detection of a specific posture can be reduced.
  • the setting values such as the notification threshold value and the weighting may be different depending on the circumstances of the piggery, the type of pigs housed in each, and the like.
  • the record up to that point is reset every time the reference observation time elapses, and counting is started from 0 at the new reference observation time, but the counting method is not limited to this.
  • the reference observation time for example, 24 hours
  • the observation result for the oldest unit time at the time of starting the observation for a new unit time may be discarded.
  • the output destination of the excess notification is the keeper terminal 220 or the notification light 240, but the output destination is not limited to these.
  • the counting unit 113 may directly display the information regarding the excess notification on the display monitor 150 connected to the server 100. Further, the counting unit 113 constantly counts the number of specific postures during counting in addition to the case where the excess notification is output when the number of times of the specific posture to be counted exceeds the notification threshold value, or instead of outputting the excess notification. It may be output. For example, the number of times of the current specific posture counted by each pen 301 may be displayed in a list on the keeper terminal 220.
  • the number of times is counted for four specific postures: the dog sitting posture, the eaves posture, the back bay posture, and the punting, but the target specific posture is limited to this. do not have.
  • One of the four may be selected, and other specific postures indicated by the unwell pig may be added to or replaced by these.
  • the server 100 functions as a pig breeding support device
  • the hardware configuration is not limited to this. If the mobile terminal described as the keeper terminal 220 performs the same processing as the server 100, the mobile terminal can function as a pig breeding support device. Further, for example, if the keeper terminal 220 is configured to take part in the processing of the server 100, the system in which the server 100 and the keeper terminal 220 cooperate with each other can be a pig breeding support device.

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Environmental Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Animal Husbandry (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Artificial Intelligence (AREA)
  • Image Analysis (AREA)

Abstract

This pig rearing assistance apparatus is provided with: an acquisition unit that acquires image data of an image captured by a camera directed to a plurality of pens in each of which a group of pigs are being raised; a detection unit that detects a specific posture of the pigs on the basis of the image of the image data; and a count unit that counts the number of times of the specific posture detected for each of the plurality of pens during a preset observation time. This pig rearing assistance apparatus can notify rearing staff of which pen includes a pig having a poor health condition among the plurality of pens in each of which the group of pigs are being raised at an appropriate timing without excessive cost and labor.

Description

豚飼育支援装置、豚飼育支援方法、および豚飼育支援プログラムPig breeding support equipment, pig breeding support methods, and pig breeding support programs
　本発明は、豚飼育支援装置、豚飼育支援方法、および豚飼育支援プログラムに関する。 The present invention relates to a pig breeding support device, a pig breeding support method, and a pig breeding support program.
　家畜の異常を検知するシステムが知られている。例えば、個々の家畜に各種センサを装着し、その出力が健康状態ではあり得ない値である場合に、当該個体を異常であると判断する（例えば、特許文献１参照）。 A system for detecting abnormalities in livestock is known. For example, when various sensors are attached to individual livestock and the output is a value that cannot be in a healthy state, the individual is judged to be abnormal (see, for example, Patent Document 1).
特開２０１７－２０１９３０号公報JP-A-2017-201930
　様々な家畜に対して発生し得る様々な異常を検知しようとすると、個々の家畜に複数のセンサを装着したり、膨大な解析データを用意したりする必要がある。豚を飼育する場合には、一般的にペンと呼ばれる檻や区切られた区画において集団飼育する手法が採用されることが多く、異常を検出するためのセンサを一頭一頭に取り付けることは現実的ではない。また、そのための膨大な解析データを用意することも難しい。さらに、養豚場の豚舎内には複数のペンが設置されることが多く、飼育員が個々の豚を万遍なく監視し続けることも困難である。 In order to detect various abnormalities that can occur in various livestock, it is necessary to attach multiple sensors to each livestock and prepare a huge amount of analysis data. When breeding pigs, a method of group breeding is often adopted in a cage or a divided section called a pen, and it is not realistic to attach a sensor for detecting abnormalities to each pig. do not have. It is also difficult to prepare a huge amount of analysis data for that purpose. In addition, multiple pens are often installed in pig farms, making it difficult for zookeepers to keep an eye on individual pigs.
　本発明は、このような問題を解決するためになされたものであり、いずれのペンに体調不良を起こしている豚が存在するかを、過度なコストや労力を要することなく適時に飼育員に知らせることのできる豚飼育支援装置等を提供するものである。 The present invention has been made to solve such a problem, and it is possible to tell the keeper in a timely manner which pen has a sick pig without undue cost and labor. It provides a pig breeding support device and the like that can inform.
　本発明の第１の態様における豚飼育支援装置は、それぞれにおいて豚が集団飼育されている複数のペンに向けられて設置されたカメラによって撮像された画像の画像データを取得する取得部と、画像データの画像に基づいて豚の特定姿勢を検知する検知部と、設定された観察時間の間に複数のペンのそれぞれにおいて検知された特定姿勢の回数を計数する計数部とを備える。 The pig breeding support device according to the first aspect of the present invention includes an acquisition unit that acquires image data of an image captured by a camera installed toward a plurality of pens in which pigs are group-reared, and an image. It includes a detection unit that detects a specific posture of a pig based on an image of data, and a counting unit that counts the number of specific postures detected by each of a plurality of pens during a set observation time.
　また、本発明の第２の態様における豚飼育支援方法は、それぞれにおいて豚が集団飼育されている複数のペンに向けられて設置されたカメラによって撮像された画像の画像データを取得する取得ステップと、画像データの画像に基づいて豚の特定姿勢を検知する検知ステップと、設定された観察時間の間に複数のペンのそれぞれにおいて検知された特定姿勢の回数を計数する計数ステップとを有する。 Further, the pig breeding support method according to the second aspect of the present invention includes an acquisition step of acquiring image data of an image captured by a camera installed toward a plurality of pens in which pigs are group-reared in each. It has a detection step of detecting a specific posture of a pig based on an image of image data, and a counting step of counting the number of times of the specific posture detected by each of a plurality of pens during a set observation time.
　また、本発明の第３の態様における豚飼育支援プログラムは、それぞれにおいて豚が集団飼育されている複数のペンに向けられて設置されたカメラによって撮像された画像の画像データを取得する取得ステップと、画像データの画像に基づいて豚の特定姿勢を検知する検知ステップと、設定された観察時間の間に複数のペンのそれぞれにおいて検知された特定姿勢の回数を計数する計数ステップとをコンピュータに実行させる。 Further, the pig breeding support program according to the third aspect of the present invention includes an acquisition step of acquiring image data of an image captured by a camera installed toward a plurality of pens in which pigs are group-reared in each. , A detection step that detects a specific posture of a pig based on an image of image data and a counting step that counts the number of specific postures detected by each of a plurality of pens during a set observation time are executed on a computer. Let me.
　本発明により、それぞれにおいて豚が集団飼育されている複数のペンのうちいずれのペンに体調不良を起こしている豚が存在するかを、豚の習性を利用することにより、過度なコストや労力を要することなく適時に飼育員に知らせることのできる豚飼育支援装置等を提供することができる。 According to the present invention, it is possible to determine which of the plurality of pens in which pigs are bred in a group has a pig that is in poor physical condition by utilizing the habit of the pig, thereby causing excessive cost and labor. It is possible to provide a pig breeding support device or the like that can notify the keeper in a timely manner without needing it.
本実施形態に係る豚飼育支援装置を採用した養豚環境の全体像を示す図である。It is a figure which shows the whole image of the pig farming environment which adopted the pig breeding support device which concerns on this embodiment. 特定姿勢の例を説明する図である。It is a figure explaining the example of a specific posture. 学習モデルを用いた特定姿勢の検知処理の手順を説明する図である。It is a figure explaining the procedure of the detection process of a specific posture using a learning model. 豚飼育支援装置と周辺装置のハードウェア構成を示す図である。It is a figure which shows the hardware composition of a pig breeding support device and a peripheral device. 基準観察時間を説明する図である。It is a figure explaining the reference observation time. 計数リストを説明する図である。It is a figure explaining the counting list. 超過通知を受けた飼育員端末の表示例を示す図である。It is a figure which shows the display example of the keeper terminal which received the excess notification. 演算部の処理手順を説明するフロー図である。It is a flow figure explaining the processing procedure of a calculation part. 他の実施例に係る豚飼育支援装置の計数リストを説明する図である。It is a figure explaining the counting list of the pig breeding support apparatus which concerns on other Examples. 更に他の実施例に係る豚飼育支援装置を採用した養豚環境の全体像を示す図である。It is a figure which shows the whole picture of the pig farming environment which adopted the pig breeding support device which concerns on another Example. 更に他の実施例に係る豚飼育支援装置の計数リストを説明する図である。It is a figure explaining the counting list of the pig breeding support apparatus which concerns on still another Examples.
　以下、発明の実施の形態を通じて本発明を説明するが、特許請求の範囲に係る発明を以下の実施形態に限定するものではない。また、実施形態で説明する構成の全てが課題を解決するための手段として必須であるとは限らない。 Hereinafter, the present invention will be described through embodiments of the invention, but the invention according to the claims is not limited to the following embodiments. Moreover, not all of the configurations described in the embodiments are indispensable as means for solving the problem.
　図１は、本実施形態に係る豚飼育支援装置を採用した養豚環境の全体像を示す図である。養豚場は、壁や柵によって区分された複数のペン３０１を備える。それぞれのペン３０１には複数（例えば１０頭程度）の豚３０２が収容され、集団で飼育されている。なお、それぞれのペン３０１で飼育される豚３０２の頭数は、豚３０２の品種や飼育環境等に応じて調整され得る。 FIG. 1 is a diagram showing an overall picture of a pig farming environment that employs the pig breeding support device according to the present embodiment. The pig farm is equipped with a plurality of pens 301 separated by walls and fences. Each pen 301 accommodates a plurality of pigs 302 (for example, about 10 pigs) and is bred in a group. The number of pigs 302 raised by each pen 301 can be adjusted according to the breed of pig 302, the breeding environment, and the like.
　収容された豚３０２を観察するためのカメラユニット２１０が、ペン３０１ごとに設置されている。カメラユニット２１０は、観察対象であるペン３０１の全体を俯瞰して撮像できるように当該ペン３０１に向けて、例えば天井から吊り下げられて設置されている。カメラユニット２１０は、撮像した画像を画像データに変換し、ネットワーク２００を介してサーバ１００へ送信する。具体的には、施設内に設置された無線ユニット２３０がネットワーク２００と接続されており、カメラユニット２１０は、無線ユニット２３０と無線通信を確立することにより、画像データをサーバ１００へ送信することができる。もちろん、有線接続によって通信を確立するように構成してもよい。なお、カメラユニット２１０とサーバ１００を接続するネットワーク２００は、インターネットやイントラネットを用いてもよいし、サーバ１００が設置される管理施設が養豚場内に設けられるような場合には、近距離無線通信を採用してもよい。 A camera unit 210 for observing the housed pig 302 is installed for each pen 301. The camera unit 210 is installed so as to be suspended from the ceiling, for example, toward the pen 301 so that the entire pen 301 to be observed can be viewed from a bird's-eye view. The camera unit 210 converts the captured image into image data and transmits the captured image to the server 100 via the network 200. Specifically, the wireless unit 230 installed in 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. Of course, it may be configured to establish communication by a wired connection. The network 200 connecting the camera unit 210 and the server 100 may use the Internet or an intranet, and when the management facility where the server 100 is installed is provided in the pig farm, short-range wireless communication is performed. It may be adopted.
　豚３０２を世話する飼育員は、飼育員端末２２０を所持し得る。飼育員端末２２０は、例えば、タブレット端末やスマートフォンであり、無線ユニット２３０およびネットワーク２００を介してサーバ１００との間で各種情報の授受を行うことができる。飼育員は、例えば飼育記録を飼育員端末２２０へ入力してサーバ１００へ転送することもできるし、サーバ１００に蓄積された飼育記録を呼び出すこともできる。また、飼育員端末２２０は、後述する超過通知を受け取った場合には、その旨を表示する。 The keeper who takes care of the pig 302 can carry the keeper terminal 220. The keeper terminal 220 is, for example, a tablet terminal or a smartphone, and can exchange various information with and from the server 100 via the wireless unit 230 and the network 200. For example, the keeper can input the breeding record to the keeper terminal 220 and transfer it to the server 100, or can call up the breeding record stored in the server 100. In addition, when the keeper terminal 220 receives the excess notification described later, it displays that fact.
　管理施設には、豚飼育支援装置としてのサーバ１００が設置されている。サーバ１００は、ネットワーク２００と接続されている。サーバ１００は、ペン３０１ごとに設置されているカメラユニット２１０から順次画像データを取得して、当該画像データに基づいて豚３０２の特定姿勢を検知し、その回数を計数する。特定姿勢については後述する。サーバ１００は、管理者や飼育員の要求に応じてその結果を表示モニタ１５０へ表示することができる。表示モニタ１５０は、例えば液晶パネルを備えるモニタである。また、サーバ１００は、管理者や飼育員の操作を受け付ける入力デバイス１６０と接続される。入力デバイス１６０は、キーボード、マウス、表示モニタ１５０の表示面に重畳されたタッチパネル等によって構成される。 A server 100 as a pig breeding support device is installed in the management facility. The server 100 is connected to the network 200. The server 100 sequentially acquires image data from the camera unit 210 installed for each pen 301, detects a specific posture of the pig 302 based on the image data, and counts the number of times. The specific posture will be described later. The server 100 can display the result on the display monitor 150 in response to a request from the administrator or the keeper. The display monitor 150 is, for example, a monitor including a liquid crystal panel. Further, the server 100 is connected to an input device 160 that accepts operations by an administrator or a keeper. The input device 160 is composed of a keyboard, a mouse, a touch panel superimposed on the display surface of the display monitor 150, and the like.
　さて、養豚において、豚３０２の体調不良の見極めは重要である。特定の豚が体調不良の兆候を見せた場合には、個別に治療にあたる必要があり、ペン３０１内の多くの豚に体調不良の傾向が見られる場合には、飼料や飲料水に薬剤を混ぜたり環境を調整したりする必要がある。しかし、このように集団飼育されているペンが豚舎内に複数設置されているような比較的規模の大きな養豚においては、飼育員が多数の豚を定常的に観察してその体調不良を見極めることは、飼育員の多大な労力や熟練を要する作業であった。本実施形態における豚飼育支援装置は、ペン３０１内の豚３０２が体調不良時に示す傾向のある特定の姿勢を検知することにより、体調不良である可能性が高い豚が少なくとも１頭以上現れた特定ペンを飼育員へ知らせることにより飼育作業を支援する。 By the way, in pig farming, it is important to identify the poor physical condition of pig 302. If a particular pig shows signs of illness, it should be treated individually, and if many pigs in Pen 301 tend to be ill, mix the drug with feed or drinking water. And the environment needs to be adjusted. However, in a relatively large-scale pig farm where multiple pens that are group-reared in this way are installed in the piggery, the keeper should constantly observe a large number of pigs to determine their physical condition. Was a task that required a great deal of labor and skill of the keeper. The pig breeding support device in the present embodiment detects at least one pig that is likely to be in poor physical condition by detecting a specific posture that the pig 302 in the pen 301 tends to show when it is in poor physical condition. Support the breeding work by notifying the keeper of the pen.
　図１に示すように養豚場に複数のペン３０１が存在する場合には、飼育員は、どのペン３０１に体調不良の兆候を示す豚３０２が含まれるかの情報を豚飼育支援装置から知らされるだけでも、多くの労力を削減することができる。すなわち、飼育員は、体調不良の兆候を示す豚３０２の抽出作業を、豚飼育支援装置から通知されるまでは行わなくてもよく、通知された場合もその特定ペンを対象として行えばよい。一つのペン内の限られた頭数に対して体調不良の兆候を示す豚３０２の存在が示されているのであれば、経験の浅い飼育員であっても、その抽出は比較的容易である。 When there are a plurality of pens 301 in the pig farm as shown in FIG. 1, the keeper is informed by the pig breeding support device of information on which pen 301 contains the pig 302 showing a sign of poor physical condition. You can save a lot of effort just by doing this. That is, the keeper does not have to perform the extraction work of the pig 302 showing a sign of poor physical condition until notified by the pig breeding support device, and even when notified, the keeper may perform the extraction work for the specific pen. If the presence of pigs 302, which show signs of poor health for a limited number of pigs in a pen, is shown, even inexperienced breeders are relatively easy to extract.
　本実施形態における豚飼育支援装置は、カメラユニット２１０で撮像された画像を利用して豚３０２の特定姿勢を検知する。特定姿勢は、豚が体調不良時に示す傾向のある姿勢であり、出願人は、その具体的な姿勢のいくつかを知見として得ている。図２は、特定姿勢の例を説明する図である。 The pig breeding support device in the present embodiment detects a specific posture of the pig 302 by using an image captured by the camera unit 210. The specific posture is the posture that the pig tends to show when he / she is in poor physical condition, and the applicant has obtained some of the specific postures as knowledge. FIG. 2 is a diagram illustrating an example of a specific posture.
　図２（Ａ）は、犬座姿勢であり、具体的には前脚を立てた状態で着座する姿勢である。後脚は折り曲げられて腹部に接していることが多く、頭部は持ち上げられている時もあれば、垂れている時もある。犬座姿勢は、豚が体調不良時に示す傾向のある姿勢の一つであり、本実施形態においては特定姿勢として扱う。 FIG. 2A shows a dog-sitting posture, specifically, a sitting posture with the front legs upright. The hind legs are often bent and in contact with the abdomen, and the head may be lifted or hung down. The dog-sitting posture is one of the postures that pigs tend to show when they are in poor physical condition, and is treated as a specific posture in this embodiment.
　図２（Ｂ）は、庇寝姿勢であり、具体的には伏臥した胴部の下に前脚を収める姿勢である。すなわち、前脚を折りたたんで、床面と自身の胴部の間で挟み込むような姿勢であり、頭部も床面に接しておとなしくしている場合が多い。庇寝姿勢は、豚が体調不良時に示す傾向のある姿勢の一つであり、本実施形態においては特定姿勢として扱う。 FIG. 2B shows a dozing posture, specifically, a posture in which the front legs are placed under the prone torso. That is, the posture is such that the front legs are folded and sandwiched between the floor surface and its own torso, and the head is often in contact with the floor surface and is quiet. The eaves posture is one of the postures that the pig tends to show when he / she is in poor physical condition, and is treated as a specific posture in this embodiment.
　図２（Ｃ）は、背湾姿勢であり、具体的には背中を丸めて前屈みとなる姿勢である。このとき、前後の脚先間の間隔は、脚元間の間隔よりも狭くなる。背湾姿勢は、豚が体調不良時に示す傾向のある姿勢の一つであり、本実施形態においては特定姿勢として扱う。 Fig. 2 (C) shows the back bay posture, specifically, the posture in which the back is curled up and leans forward. At this time, the distance between the front and rear leg tips is narrower than the distance between the leg bases. The back bay posture is one of the postures that pigs tend to show when they are in poor physical condition, and is treated as a specific posture in this embodiment.
　図２（Ｄ）は、パンティングであり、具体的には横臥位で腹部を波打たせる、動作を伴う姿勢である。横臥位自体は健常な豚にも見られる姿勢であるが、体調不良の豚は腹部を繰り返し上下させて床面に打ち付ける動作を見せることがある。パンティングは、豚が体調不良時に示す傾向のある動作付随姿勢の一つであり、本実施形態においては特定姿勢として扱う。 FIG. 2 (D) is punting, specifically, a posture accompanied by movement, in which the abdomen is wavy in the lying position. The lying position itself is a posture that is also seen in healthy pigs, but unwell pigs may show the movement of repeatedly raising and lowering the abdomen and hitting it against the floor. Punting is one of the movement-associated postures that pigs tend to show when they are in poor physical condition, and is treated as a specific posture in this embodiment.
　他の家畜として、例えば牛についても体調不良の兆候を健康時とは異なる姿勢の有無によって推定することも考えられる。しかし、牛は、少数で飼育されていたり、一頭一頭の価値が高かったりすることから、個々の牛に姿勢を検知する例えば赤外線センサや加速度センサを装着する手法が採用される。しかし、限られたスペースのペンに比較的多くが収容される豚の集団飼育において、個々の豚にセンサを装着することは現実的ではない。特に、一頭当たりの豚の出荷価格は牛の出荷価格に比べて安価であるため、飼育にコストを掛けることが難しい。 For other livestock, for example, cattle, it is possible to estimate the signs of poor physical condition based on the presence or absence of a posture different from that at the time of health. However, since cows are bred in small numbers or the value of each cow is high, a method of attaching, for example, an infrared sensor or an acceleration sensor to detect the posture of each cow is adopted. However, in a group breeding of pigs in which a relatively large number of pigs are housed in a pen with a limited space, it is not realistic to attach a sensor to each pig. In particular, since the shipping price of pigs per pig is lower than the shipping price of cattle, it is difficult to increase the cost of breeding.
　このような背景を鑑み、本実施形態に係る豚飼育支援装置は、ペン３０１に向けて設置されたカメラユニット２１０を用いて撮像された画像に基づいて集団飼育されている豚３０２の特定姿勢を検知する。なお、それぞれのカメラユニット２１０は、撮像対象のペン３０１内に収容された豚３０２の姿勢が捉えられるように、傾斜方向へ見下ろすように設置されている。 In view of such a background, the pig breeding support device according to the present embodiment determines the specific posture of the pig 302 group-reared based on the image captured by the camera unit 210 installed toward the pen 301. Detect. Each camera unit 210 is installed so as to look down in the inclined direction so that the posture of the pig 302 housed in the pen 301 to be imaged can be captured.
　ペン３０１を俯瞰する画像からは、例えば、一頭ごとの輪郭線を抽出し、それぞれの輪郭線の形状をテンプレートパターンとパターンマッチングすることにより、特定姿勢を検知することができる。テンプレートパターンは、それぞれの姿勢を様々な角度から捉えた輪郭線であり、予め用意される。なお、具体的には後述するが、動きの変化がなくても検知できる犬座姿勢、庇寝姿勢、背湾姿勢は静止画の単画像から検知できるが、動作付随姿勢であるパンティングは前後する複数枚の静止画、動画を用いて検知する。 A specific posture can be detected by, for example, extracting the contour line of each pen from the image overlooking the pen 301 and pattern matching the shape of each contour line with the template pattern. The template pattern is a contour line that captures each posture from various angles, and is prepared in advance. As will be described in detail later, the dog-sitting posture, the eaves posture, and the back bay posture, which can be detected even if there is no change in movement, can be detected from a single image of a still image, but the punting, which is a movement-accompanying posture, is front and back. Detect using multiple still images and moving images.
　このように、集団飼育する豚３０２の俯瞰画像から特定姿勢を検知する手法によれば、例えば、センサを装着する二頭が接近することによって特定姿勢であると検知されてしまう誤検知の懸念もない。また、屋外の広い牧場で集団飼育される牛等の家畜と異なり、屋内のペンで集団飼育される豚３０２に対しては、俯瞰するカメラユニット２１０の設置が比較的容易であり、光量や撮像画角を安定させやすいことから、画像を用いた検知手法は好適である。 In this way, according to the method of detecting a specific posture from a bird's-eye view image of a group-reared pig 302, for example, there is a concern of false detection that two pigs wearing sensors are detected as having a specific posture when they approach each other. do not have. In addition, unlike cattle and other livestock that are group-raised on a large outdoor ranch, it is relatively easy to install a bird's-eye view camera unit 210 for pigs 302 that are group-reared with an indoor pen, and the amount of light and imaging are possible. Since it is easy to stabilize the angle of view, the detection method using an image is suitable.
　上述のように取得した画像から輪郭線を抽出してパターンマッチングにより特定姿勢を検知することもできるが、本実施形態においては、対象とする特定姿勢を取る豚の画像を教師画像として学習した学習モデルへ取得した画像を入力して特定姿勢を検知する。図３は、学習モデルである検知用ニューラルネットワーク１２１を用いた特定姿勢の検知処理の手順を説明する図である。 It is also possible to extract a contour line from the image acquired as described above and detect a specific posture by pattern matching, but in the present embodiment, learning is performed by learning an image of a pig taking a target specific posture as a teacher image. The acquired image is input to the model to detect a specific posture. FIG. 3 is a diagram illustrating a procedure for detecting a specific posture using the detection neural network 121 which is a learning model.
　検知用ニューラルネットワーク１２１は、犬座姿勢を取る豚の画像に「犬座姿勢」を正解として紐づけた教師データ、庇寝姿勢を取る豚の画像に「庇寝姿勢」を正解として紐づけた教師データ、背湾姿勢を取る豚の画像に「背湾姿勢」を正解として紐づけた教師データ、パンティングを取る豚の画像に「パンティング」を正解として紐づけた教師データをそれぞれ相当数与えて学習させる教師あり学習によって予め作成されたものである。本実施形態においては動きの変化を伴うパンティングも特定姿勢に加えているので、検知用ニューラルネットワーク１２１の入力画像を、パンティングが検知できる程度の長さ（例えば３秒）の動画像とする。したがって、学習時の教師データもそれぞれ動画像で用意される。なお、特定姿勢として動きの変化を伴う姿勢を加えないのであれば、検知用ニューラルネットワーク１２１の入力画像を静止画としてもよい。 The detection neural network 121 associates the image of the pig in the dog-sitting posture with the teacher data as the correct answer, and the image of the pig in the dozing posture with the "sleeping posture" as the correct answer. A considerable number of teacher data, teacher data linking the image of a pig taking a back bay posture with "back bay posture" as the correct answer, and teacher data linking an image of a pig taking a punting with "punting" as the correct answer, respectively. It is created in advance by supervised learning to give and learn. In the present embodiment, punting accompanied by a change in movement is also added to the specific posture, so that the input image of the detection neural network 121 is a moving image having a length (for example, 3 seconds) that can detect punting. .. Therefore, the teacher data at the time of learning is also prepared as a moving image. The input image of the detection neural network 121 may be a still image as long as the posture accompanied by the change in movement is not added as the specific posture.
　このように生成された学習モデルである検知用ニューラルネットワーク１２１は、豚飼育支援装置であるサーバ１００に組み込まれて利用に供される。具体的には、例えば養豚場に観察対象であるペン３０１が８つ設けられているとすると、それぞれのペン３０１を俯瞰する８つのカメラユニット２１０から順次画像データがサーバ１００へ送られてくる。それらの画像データの画像ｉｍｇ１～ｉｍｇ８は、検知用ニューラルネットワーク１２１へ順次入力される。 The detection neural network 121, which is the learning model generated in this way, is incorporated into the server 100, which is a pig breeding support device, and is used. Specifically, for example, assuming that eight pens 301 to be observed are provided in the pig farm, image data is sequentially sent to the server 100 from the eight camera units 210 that overlook each pen 301. The images img 1 to img 8 of the image data are sequentially input to the detection neural network 121.
　検知用ニューラルネットワーク１２１は、画像が入力されるごとに、その画像内で特定姿勢と判定される豚の頭数を検知数として出力する。具体的には、当該画像内に犬座姿勢を取る頭数、庇寝姿勢を取る頭数、背湾姿勢を取る頭数、パンティングを取る頭数をそれぞれの姿勢の検知数として出力する。このようにして、例えば第１ペンの画像ｉｍｇ１に対する検知数は、犬座姿勢「０」、庇寝姿勢「１」、背湾姿勢「１」、パンティング「１」、第２ペンの画像ｉｍｇ２に対する検知数は、犬座姿勢「２」、庇寝姿勢「０」、背湾姿勢「０」、パンティング「０」…のように、各ペン３０１に対するそれぞれの姿勢の検知数を得る。サーバ１００は、このような処理を予め設定された観察時間の間繰り返し、ペン３０１ごとに検知数を計数・積算する。 Each time an image is input, the detection neural network 121 outputs the number of pigs determined to have a specific posture in the image as the number of detections. Specifically, the number of dogs in the dog-sitting posture, the number of heads in the eaves posture, the number of heads in the dorsal bay posture, and the number of heads in the punting posture are output as the detected number of each posture in the image. In this way, for example, the number of detections for the image img 1 of the first pen is the dog sitting posture "0", the dozing posture "1", the back bay posture "1", the punting "1", and the image of the second pen. As for the number of detections for img 2 , the number of detections of each posture for each pen 301 is obtained, such as the dog sitting posture "2", the dozing posture "0", the back bay posture "0", and the punting "0". .. The server 100 repeats such a process for a preset observation time, and counts and integrates the number of detections for each pen 301.
　図４は、豚飼育支援装置としてのサーバ１００と周辺装置のハードウェア構成を示す図である。サーバ１００は、上述のように、表示モニタ１５０、入力デバイス１６０、カメラユニット２１０、飼育員端末２２０と接続可能である。 FIG. 4 is a diagram showing the hardware configuration of the server 100 as a pig breeding support device and peripheral devices. As described above, the server 100 can be connected to the display monitor 150, the input device 160, the camera unit 210, and the keeper terminal 220.
　サーバ１００は、主に、演算部１１０、記憶部１２０、通信ユニット１３０によって構成される。演算部１１０は、サーバ１００の制御とプログラムの実行処理を行うプロセッサ（ＣＰＵ：Central Processing Unit）である。プロセッサは、ＡＳＩＣ（Application Specific Integrated Circuit）やＧＰＵ（Graphics Processing Unit）等の演算処理チップと連携する構成であってもよい。演算部１１０は、記憶部１２０に記憶された豚飼育支援プログラムを読み出して、豚飼育の支援に関する様々な処理を実行する。 The server 100 is mainly composed of a calculation unit 110, a storage unit 120, and a communication unit 130. The arithmetic unit 110 is a processor (CPU: Central Processing Unit) that controls the server 100 and executes a program. 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 breeding support program stored in the storage unit 120 and executes various processes related to the pig breeding support.
　記憶部１２０は、不揮発性の記憶媒体であり、例えばＨＤＤ（Hard Disk Drive）によって構成されている。記憶部１２０は、サーバ１００の制御や処理を実行するプログラムの他にも、制御や演算に用いられる様々なパラメータ値、関数、表示要素データ、ルックアップテーブル等を記憶し得る。記憶部１２０は、特に、検知用ニューラルネットワーク１２１と計数リスト１２２を記憶している。検知用ニューラルネットワーク１２１は、上述のように、カメラユニット２１０が撮像した画像を入力すると、その画像内に存在する特定姿勢の頭数を表す検知数を出力する。計数リスト１２２は、特定姿勢の計数に関する記録であり、具体的には後述する。なお、記憶部１２０は、複数のハードウェアで構成されていても良く、例えば、プログラムを記憶する記憶媒体と検知用ニューラルネットワーク１２１を記憶する記憶媒体が別々のハードウェアで構成されてもよい。 The storage unit 120 is a non-volatile storage medium, and is composed of, for example, an HDD (Hard Disk Drive). The storage unit 120 can store various parameter values, functions, display element data, lookup tables, and the like used for control and calculation, in addition to a program that executes control and processing of the server 100. The storage unit 120 stores, in particular, the detection neural network 121 and the counting list 122. As described above, when the image captured by the camera unit 210 is input, the detection neural network 121 outputs the number of detections representing the number of heads of a specific posture existing in the image. The counting list 122 is a record relating to counting of a specific posture, and will be described in detail later. The storage unit 120 may be composed of a plurality of hardware. For example, the storage medium for storing the program and the storage medium for storing the detection neural network 121 may be composed of different hardware.
　通信ユニット１３０は、例えばＬＡＮユニットを含み、ネットワーク２００を介して、演算部１１０が生成する撮像制御信号をカメラユニット２１０へ送信したり、カメラユニット２１０から送られてくる画像データを演算部１１０へ引き渡したりする。また、飼育員端末２２０と演算部１１０の間で実行されるデータの授受を中継する。なお、通信ユニット１３０は、他の外部装置との間でデータや制御信号の授受を中継することもできる。例えば、豚飼育支援プログラムや検知用ニューラルネットワーク１２１の更新データを外部サーバから取り込む場合にも利用され得る。 The communication unit 130 includes, for example, a LAN unit, transmits an imaging control signal generated by the calculation unit 110 to the camera unit 210, and transmits image data sent from the camera unit 210 to the calculation unit 110 via the network 200. I will hand it over. In addition, the transfer of data executed between the keeper terminal 220 and the calculation unit 110 is relayed. The communication unit 130 can also relay the exchange of data and control signals with other external devices. For example, it can also be used when fetching update data of a pig breeding support program or a detection neural network 121 from an external server.
　演算部１１０は、豚飼育支援プログラムが指示する処理に応じて様々な演算を実行する機能演算部としての役割も担う。演算部１１０は、取得部１１１、検知部１１２、計数部１１３として機能し得る。取得部１１１は、主に、カメラユニット２１０によって撮像された画像の画像データを取得し、検知部１１２へ引き渡す。検知部１１２は、主に、取得部１１１から受け取った画像データの画像に基づいて豚３０２の特定姿勢を検知し、その結果を計数部１１３へ引き渡す。計数部１１３は、主に、設定された観察時間の間に複数のペン３０１のそれぞれにおいて検知された特定姿勢の回数を計数する。 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, and a counting unit 113. The acquisition unit 111 mainly acquires the image data of the image captured by the camera unit 210 and delivers it to the detection unit 112. The detection unit 112 mainly detects the specific posture of the pig 302 based on the image of the image data received from the acquisition unit 111, and delivers the result to the counting unit 113. The counting unit 113 mainly counts the number of times of the specific posture detected by each of the plurality of pens 301 during the set observation time.
　次に、計数部１１３の処理について説明する。図５は、計数部１１３が特定姿勢の回数を計数する対象期間としての基準観察時間を説明する図である。計数部１１３は、予め設定された基準観察時間ごとに計数した回数をリセットする。 Next, the processing of the counting unit 113 will be described. FIG. 5 is a diagram for explaining a reference observation time as a target period for the counting unit 113 to count the number of times of a specific posture. The counting unit 113 resets the number of times counted for each preset reference observation time.
　本実施形態においては、基準観察時間を２４時間に設定している。計数部１１３は、この基準観察時間の間に各ペン３０１において何回の特定姿勢が検知されたかを計数する。具体的には、図３を用いて説明したように、取得部１１１が周期的に取得する各ペン３０１の画像データの画像ｉｍｇ１～ｉｍｇ８を、検知部１１２が検知用ニューラルネットワーク１２１を用いてそれぞれの画像内で特定姿勢を取る頭数（検知数）を検知する。これを基準観察時間の間繰り返し、各ペン３０１において何回の特定姿勢が検知されたかを数え上げる。 In this embodiment, the reference observation time is set to 24 hours. The counting unit 113 counts how many times a specific posture is detected in each pen 301 during this reference observation time. Specifically, as described with reference to FIG. 3, the detection unit 112 uses the detection neural network 121 to use the images img 1 to img 8 of the image data of each pen 301 periodically acquired by the acquisition unit 111. The number of heads (detected number) that take a specific posture in each image is detected. This is repeated during the reference observation time, and the number of times a specific posture is detected by each pen 301 is counted.
　ただし、基準観察時間に含まれる期間であっても、一部の状況に対応する期間については計数の対象から除外する。ここではそのような除外期間として、飼育員接近期間が設定されている。飼育員接近期間は、ペン３０１へ飼育員が接近してから離隔するまでの予定時間を含む一定期間として設定されている。飼育員接近期間は、例えば給餌のために飼育員がペン３０１内に進入する８時から９時が設定されている。この他にも、ペン３０１の清掃や豚３０２の検査等のために飼育員が接近する期間も飼育員接近期間として設定してもよい。 However, even if the period is included in the standard observation time, the period corresponding to some situations is excluded from the counting. Here, a keeper approach period is set as such an exclusion period. The keeper approach period is set as a fixed period including a scheduled time from when the keeper approaches the pen 301 to when the keeper approaches the pen 301. The keeper approach period is set from 8:00 to 9:00 when the keeper enters the pen 301 for feeding, for example. In addition to this, a period during which the keeper approaches for cleaning the pen 301, inspecting the pig 302, or the like may also be set as the keeper approach period.
　このように、飼育員がペン３０１に接近したり進入したりすると、豚３０２は一定の興奮状態になり、健康な豚が特定姿勢を取ったり、体調不良の豚が特定姿勢を崩したりすることがある。したがって、このような不規則な行動を取る可能性のあるイベントに対しては、そのイベントの実施期間を含む一定期間を計数の対象から除外する除外期間とするとよい。具体的には、除外期間は、イベントの実施期間に前後する、豚３０２がイベントを察知して興奮しだす期間、およびイベント後に落ち着きを取り戻す期間を付加して設定されるとよい。 In this way, when the keeper approaches or enters the pen 301, the pig 302 becomes a certain state of excitement, and a healthy pig takes a specific posture or a poorly ill pig loses a specific posture. There is. Therefore, for an event that may behave irregularly like this, it is advisable to set an exclusion period that excludes a certain period including the implementation period of the event from the target of counting. Specifically, the exclusion period may be set by adding a period before and after the event implementation period, a period in which the pig 302 senses the event and begins to get excited, and a period in which the pig 302 regains calm after the event.
　なお、除外期間は、管理者等が入力デバイス１６０を操作して事前に設定することができる。除外期間は、毎回の基準観察時間に含まれるものでなくてもよく、例えば雌豚を飼育するペンへ雄豚を短時間の間投入するイベントであれば、そのイベントを行う特定日における基準観察時間に対してのみ設定されてもよい。計数部１１３は、このように設定された除外期間については、計数の対象から除外する。具体的には、計数部１１３による計数を停止する、検知部１１２に特定姿勢の検知を停止させる、取得部１１１に画像データの取得を停止させる、カメラユニット２１０に撮像を停止させる等の処理を採用し得る。 The exclusion period can be set in advance by the administrator or the like by operating the input device 160. The exclusion period does not have to be included in each reference observation time. For example, if it is an event in which a boar is put into a pen for raising a sow for a short period of time, the reference observation on a specific day when the event is held It may be set only for time. The counting unit 113 excludes the exclusion period set in this way from the target of counting. Specifically, processing such as stopping counting by the counting unit 113, stopping the detection of a specific posture by the detection unit 112, stopping the acquisition of image data by the acquisition unit 111, and stopping the imaging by the camera unit 210 is performed. Can be adopted.
　また、検知部１１２が画像データを用いて特定姿勢を検知する周期は、豚３０２が特定姿勢を続ける標準的な継続時間等を考慮して、１回の特定姿勢が１回として検知されるように調整、設定される。周期の設定は、その他にも豚３０２の品種や飼育環境、月齢などを考慮することもできる。 Further, in the cycle in which the detection unit 112 detects the specific posture using the image data, one specific posture is detected as one time in consideration of the standard duration of the pig 302 continuing the specific posture. Adjusted and set to. In addition, the breed of pig 302, the breeding environment, the age of the moon, and the like can be taken into consideration when setting the cycle.
　計数部１１３は、ペン３０１ごとに数え上げる特定姿勢の回数を計数リスト１２２により管理する。図６は、計数リスト１２２の一例を説明する図である。計数リスト１２２は、一度の基準観察時間に一つ生成され、その観察時間の間は適宜更新される。 The counting unit 113 manages the number of times of the specific posture counted for each pen 301 by the counting list 122. FIG. 6 is a diagram illustrating an example of the counting list 122. One counting list 122 is generated at one reference observation time, and is updated as appropriate during the observation time.
　計数リスト１２２は、通知閾値、観察日を含む。通知閾値は、管理者等によって予め設定される値であり、それぞれのペン３０１において計数された特定姿勢の総回数がこの通知閾値を超えると、計数部１１３は、飼育員端末２２０等へその旨を知らせる超過通知を出力する。管理者等は、豚３０２の品種や飼育環境、特に１つのペン３０１に収容されている豚３０２の頭数などを考慮して通知閾値を設定する。より具体的には、対象となるペン３０１に体調不良の兆候を示す豚３０２が少なくとも１頭以上存在すると判定し得る閾値を、それまでの統計や経験に基づいて設定する。図６の例では、「５０回」が設定されている。観察日は、観察を実行している日付であり、観察後に計数リスト１２２が参照される場合には、観察を実行した日付を表している。基準観察時間が２４時間未満であれば、例えば観察を開始した観察時刻を加えてもよい。 The counting list 122 includes the notification threshold value and the observation date. The notification threshold value is a value preset by the administrator or the like, and when the total number of times of the specific posture counted by each pen 301 exceeds this notification threshold value, the counting unit 113 informs the keeper terminal 220 or the like to that effect. Output an excess notification to notify. The manager or the like sets the notification threshold value in consideration of the breed and breeding environment of the pig 302, particularly the number of pigs 302 housed in one pen 301. More specifically, a threshold value at which it can be determined that at least one pig 302 showing a sign of poor physical condition exists in the target pen 301 is set based on the statistics and experience so far. In the example of FIG. 6, "50 times" is set. The observation date is the date on which the observation is performed, and when the counting list 122 is referred to after the observation, it represents the date on which the observation was performed. If the reference observation time is less than 24 hours, for example, the observation time at which the observation was started may be added.
　計数リスト１２２は、ペン３０１ごとに計数した特定姿勢の回数を示す集計表を含む。集計表は、ペンナンバー（例えば、第１ペンから第８ペンまでの８つ）、特定姿勢のそれぞれ（犬座姿勢、庇寝姿勢、背湾姿勢、パンティング）について計数した回数、およびそれらを合計した総回数、フラグ情報によって構成される。計数部１１３は、計数した総回数が通知閾値を超えるペン３０１が現れたことを確認したら、超過通知を生成して飼育員端末２２０等へ出力する。このとき、計数部１１３は、計数した回数が通知閾値を超えた特定ペンに関するペン情報を超過通知に付加して出力する。本実施形態においては、ペンナンバーをペン情報として付加する。超過通知を出力したペン３０１については、フラグ情報として「通知済み」の情報が記録される。 The counting list 122 includes a tabulation table showing the number of times of a specific posture counted for each pen 301. The tabulation table shows the number of times the pen numbers (for example, 8 from the 1st pen to the 8th pen), the number of times each of the specific postures (dog sitting posture, eaves posture, back bay posture, punting) were counted, and those. It consists of the total number of times and flag information. When the counting unit 113 confirms that the pen 301 whose total number of times counted exceeds the notification threshold has appeared, it generates an excess notification and outputs it to the keeper terminal 220 or the like. At this time, the counting unit 113 adds pen information regarding the specific pen whose number of times counted exceeds the notification threshold value to the excess notification and outputs the pen information. In this embodiment, the pen number is added as pen information. For the pen 301 that outputs the excess notification, "notified" information is recorded as flag information.
　計数リスト１２２は、除外期間の情報を含む。具体的には、図５を用いて説明した除外期間がリスト情報として記録される。なお、除外期間が長ければその分だけ計数する対象期間が短くなるので、実際には体調不良の兆候を示す豚が存在する場合であっても、計数する回数が通知閾値を超えない場合もあり得る。そこで、計数部１１３は、設定された除外期間の合計時間が基準観察時間に占める割合を考慮して、通知閾値を自動的に修正してもよい。例えば、基準観察時間が２４時間であって、除外期間の合計が３時間である場合には、通知閾値を５０×（２４－３）／２４＝４３．７５回に修正する。この場合、計数部１１３は、計数した総回数が修正された通知閾値を超えるペン３０１が現れたことを確認したら超過通知を出力する。 The counting list 122 contains information on the exclusion period. Specifically, the exclusion period described with reference to FIG. 5 is recorded as list information. If the exclusion period is long, the target period for counting will be shortened accordingly, so even if there are pigs that actually show signs of poor physical condition, the number of countings may not exceed the notification threshold. obtain. Therefore, the counting unit 113 may automatically correct the notification threshold value in consideration of the ratio of the total time of the set exclusion period to the reference observation time. For example, when the reference observation time is 24 hours and the total exclusion period is 3 hours, the notification threshold is corrected to 50 × (24-3) / 24 = 43.75 times. In this case, the counting unit 113 outputs an excess notification when it confirms that the pen 301 whose total number of times counted exceeds the corrected notification threshold value has appeared.
　また、それぞれのペン３０１に収容される豚３０２の頭数が互いに異なるのであれば、収容されている頭数を考慮してペンごとに通知閾値を修正してもよい。例えば、一つのペンに１０頭の豚３０２を収容することを想定して通知閾値の「５０回」が設定されているのであれば、８頭の豚３０２が収容されたペン３０１に対しては、５０×（８／１０）＝４０回に修正する。この場合、計数部１１３は、８頭が収容されたペン３０１を対象として計数した総回数が４０回を超えたら超過通知を出力する。 Further, if the number of pigs 302 housed in each pen 301 is different from each other, the notification threshold value may be modified for each pen in consideration of the number of pigs housed. For example, if the notification threshold value of "50 times" is set on the assumption that one pen accommodates 10 pigs 302, the pen 301 containing 8 pigs 302 will be used. , 50 × (8/10) = 40 times. In this case, the counting unit 113 outputs an excess notification when the total number of times counted for the pen 301 containing eight heads exceeds 40 times.
　図７は、超過通知を受けた飼育員端末２２０の表示例を示す図である。上述のように、計数部１１３が超過通知を出力すると、飼育員端末２２０は、当該超過通知を受信して、その内容を表示パネルに表示する。具体的には図示するように、飼育員端末２２０は、超過通知に付加されているペン情報を参照して、特定姿勢の回数が通知閾値（規定値）を超えたペンナンバーを表示する（図の例では「第２ペン」）。また、養豚場のペン配置に関する屋内地図を保持している場合は、当該特定ペンの位置が認識されるように併せて表示する。なお、飼育員端末２２０は、このような表示を行うと共に告知音を発してもよい。 FIG. 7 is a diagram showing a display example of the keeper terminal 220 that has received the excess notification. As described above, when the counting unit 113 outputs the excess notification, the keeper terminal 220 receives the excess notification and displays the content on the display panel. Specifically, as shown in the figure, the keeper terminal 220 refers to the pen information added to the excess notification and displays the pen number in which the number of times of the specific posture exceeds the notification threshold (specified value) (FIG. In the example of, "second pen"). In addition, if an indoor map regarding the pen arrangement of the pig farm is held, it is also displayed so that the position of the specific pen is recognized. The keeper terminal 220 may make such a display and emit a notification sound.
　次に、サーバ１００を用いた豚飼育支援方法の処理手順について説明する。図８は、演算部１１０の処理手順を説明するフロー図である。フローは、基準観察時間の開始時点から開始される。なお、ここでは、除外期間に対する処理を省略して説明する。 Next, the processing procedure of the pig breeding support method using the server 100 will be described. FIG. 8 is a flow chart illustrating a processing procedure of the calculation unit 110. The flow starts from the start of the reference observation time. Here, the processing for the exclusion period will be omitted.
　計数部１１３は、観察開始にあたり初期処理としてステップＳ１０１で、経時タイマＴによる経時を開始させ、第ｎペンに収容されている豚３０２の特定姿勢の総回数をカウントするカウンタＣｎを全てリセットする。なお、ここでは養豚場内のペンの数はｍ個であり、それぞれのペンに対応するＣ１からＣｍまでのカウンタが用意されているものとする。 At the start of observation, the counting unit 113 starts aging by the aging timer T in step S101 as an initial process, and resets all the counters Cn for counting the total number of times of the specific posture of the pig 302 housed in the nth pen. .. Here, it is assumed that the number of pens in the pig farm is m, and counters from C 1 to C m corresponding to each pen are prepared.
　計数部１１３は、ステップＳ１０２へ進み、変数ｎを１にセットする。これに応じて処理対象とするカウンタをＣ１に切り替える。ステップＳ１０３へ進み、取得部１１１は、第ｎペンに向けられたカメラユニット２１０から、ｉｍｇｎの画像データを、通信ユニット１３０を介して取得する。ｎ＝１である場合には、第１ペンに向けられたカメラユニット２１０から、ｉｍｇ１の画像データを取得する。取得部１１１は、取得した画像データを検知部１１２へ引き渡す。 The counting unit 113 proceeds to step S102 and sets the variable n to 1. The counter to be processed is switched to C1 accordingly. Proceeding to step S103, the acquisition unit 111 acquires image data of img n from the camera unit 210 directed to the nth pen via the communication unit 130. When n = 1, the image data of img 1 is acquired from the camera unit 210 pointed at the first pen. The acquisition unit 111 delivers the acquired image data to the detection unit 112.
　検知部１１２は、ステップＳ１０４で、受け取った画像データの画像ｉｍｇｎを記憶部１２０から読み出した検知用ニューラルネットワーク１２１へ入力して、画像ｉｍｇｎ内で特定姿勢を取る豚３０２の頭数（検知数）を出力させる。具体的には図３を用いて説明したように、犬座姿勢の頭数、庇寝姿勢の頭数、背湾姿勢の頭数、パンティングの頭数を出力させる。検知部１１２は、それらの頭数を検知数として計数部１１３へ引き渡す。計数部１１３は、ステップＳ１０５で、計数リスト１２２を読み出し、受け取ったそれぞれの検知数を各々の姿勢に対応させて加算すると共に、その時点におけるＣｎのカウンタ値に検知部１１２から受け取った検知数の全数を加算することによりＣｎを更新する。 In step S104, the detection unit 112 inputs the image im g n of the received image data into the detection neural network 121 read from the storage unit 120, and the number of pigs 302 (detection number) taking a specific posture in the image im g n . ) Is output. Specifically, as described with reference to FIG. 3, the number of dogs in the dog-sitting posture, the number of heads in the eaves posture, the number of heads in the back bay posture, and the number of wheezing heads are output. The detection unit 112 delivers the number of these heads to the counting unit 113 as the number of detections. In step S105, the counting unit 113 reads out the counting list 122, adds each detected number received according to each posture, and adds the detected number received from the detection unit 112 to the counter value of Cn at that time. Cn is updated by adding all the numbers of.
　ステップＳ１０６へ進み、計数部１１３は、更新したＣｎの値が、通知閾値Ｃｄを超えたか否かを判断する。通知閾値Ｃｄを超えたと判断したら、ステップＳ１０７へ進み、超過通知を生成して飼育員端末２２０へ出力する。このとき、ペン情報として第ｎペンが特定ペンであることを付加する。超過通知を出力したらステップＳ１０８へ進む。計数部１１３は、ステップＳ１０６で通知閾値Ｃｄを超えていないと判断したら、ステップＳ１０７をスキップしてステップＳ１０８へ進む。 Proceeding to step S106, the counting unit 113 determines whether or not the updated value of Cn exceeds the notification threshold value Cd . If it is determined that the notification threshold value C d has been exceeded, the process proceeds to step S107 to generate an excess notification and output it to the keeper terminal 220. At this time, it is added that the nth pen is a specific pen as pen information. After outputting the excess notification, the process proceeds to step S108. If the counting unit 113 determines in step S106 that the notification threshold value C d has not been exceeded, the counting unit 113 skips step S107 and proceeds to step S108.
　計数部１１３は、ステップＳ１０８で変数ｎをインクリメントし、ステップＳ１０９へ進む。ステップＳ１０９へ進んだら、変数ｎが養豚場内のペン数ｍを超えたか否かを判断する。超えていないと判断したら、ステップＳ１０３へ戻り、インクリメントした変数ｎに対して同様の処理を実行する。超えたと判断したら、ステップＳ１１０へ進む。 The counting unit 113 increments the variable n in step S108, and proceeds to step S109. When the process proceeds to step S109, it is determined whether or not the variable n exceeds the number of pens m in the pig farm. If it is determined that the value is not exceeded, the process returns to step S103, and the same processing is executed for the incremented variable n. If it is determined that the value has been exceeded, the process proceeds to step S110.
　計数部１１３は、ステップＳ１１０へ進んだら、経時タイマＴが基準観察時間Ｔｃを超えたか否かを判断する。超えていないと判断したら、予め設定された周期に応じたインターバルをおいてステップＳ１０２へ戻る。超えたと判断したら、一連の処理を終了する。連続して観察を実行する場合には、再びステップＳ１０１から処理を開始する。 After proceeding to step S110, the counting unit 113 determines whether or not the time-lapse timer T has exceeded the reference observation time T c . If it is determined that the value is not exceeded, the process returns to step S102 at intervals corresponding to the preset period. When it is determined that the amount has been exceeded, a series of processes are terminated. When the observation is continuously executed, the process is started again from step S101.
　次に、本実施形態に係るいくつかの別実施例について説明する。図９は、他の実施例に係る豚飼育支援装置の計数リスト１２２’を説明する図である。計数リスト１２２’は、重み係数表を有する点で図６に示した計数リスト１２２と異なる。これまで説明した実施例においては、特定姿勢として検知の対象とした犬座姿勢、庇寝姿勢、背湾姿勢、パンティングを対等に扱い、それぞれの姿勢で検知された回数を単純に足し合わせて総回数を計数した。しかし、それぞれの姿勢が検知された場合に体調不良の兆候を示す度合い、換言すれば体調不良の判断に寄与する寄与度は、互いに異なり得る。 Next, some other embodiments relating to this embodiment will be described. FIG. 9 is a diagram illustrating a counting list 122'of a pig breeding support device according to another embodiment. The counting list 122'is different from the counting list 122 shown in FIG. 6 in that it has a weighting coefficient table. In the examples described so far, the dog sitting posture, the eaves posture, the back bay posture, and the punting, which are the targets of detection as specific postures, are treated equally, and the number of times detected in each posture is simply added up. The total number of times was counted. However, the degree to which each posture is detected to indicate a sign of poor physical condition, in other words, the degree of contribution to the judgment of poor physical condition may differ from each other.
　例えば、パンティングは、豚舎内が高温であって暑熱ストレスが高まった場合に観察されることが多く、放置すれば当該豚が死に至る可能性も高い。一方で、背湾姿勢は、軽度の体調不良の豚にもしばしば観察される。そこで、本実施例においては、特定姿勢として検知されるそれぞれの姿勢に応じて重み付けを付与して総回数を計数する。 For example, punting is often observed when the temperature inside the piggery is high and the heat stress increases, and if left untreated, the pig is likely to die. On the other hand, the back bay posture is often observed in pigs with mild illness. Therefore, in this embodiment, weighting is applied according to each posture detected as a specific posture, and the total number of times is counted.
　図９の計数リスト１２２’によれば、犬座姿勢の重み係数は「１．０」であり、庇寝姿勢の重み係数は「１．０」であり、背湾姿勢の重み係数は「０．８」であり、パンティングの重み係数は「１．６」である。ここでは、体調不良の判断に与える背湾姿勢の寄与度を下げ、パンティングの寄与度を上げている。管理者等は、経験的、統計的に把握しているそれぞれの姿勢における寄与度に応じて、事前に重み係数を設定することができる。このような重み係数が設定されると、例えば犬座姿勢が４回、庇寝姿勢が３回、背湾姿勢が４回、パンティングが１回と観察された第２ペンの総回数は、４×１．０＋３×１．０＋４×０．８＋１×１．６＝１１．８回と計算される。このように重み付けを与えて総回数をスコアのように計数することにより、それぞれの姿勢の特性を考慮した、より精度の高い体調不良予測を実現することができる。 According to the counting list 122'in FIG. 9, the weighting coefficient of the dog sitting posture is "1.0", the weighting coefficient of the dozing posture is "1.0", and the weighting coefficient of the dorsal bay posture is "0". It is "0.8" and the weighting coefficient of punting is "1.6". Here, the contribution of the back bay posture to the judgment of poor physical condition is lowered, and the contribution of punting is increased. The manager or the like can set the weighting coefficient in advance according to the degree of contribution in each posture that is empirically and statistically grasped. When such a weighting coefficient is set, for example, the total number of times of the second pen observed that the dog sitting posture is 4 times, the eaves posture is 3 times, the back bay posture is 4 times, and the punting is 1 time is calculated. It is calculated as 4 × 1.0 + 3 × 1.0 + 4 × 0.8 + 1 × 1.6 = 11.8 times. By weighting and counting the total number of times like a score in this way, it is possible to realize more accurate prediction of poor physical condition in consideration of the characteristics of each posture.
　図９の計数リスト１２２’を利用する実施例においては、検知するそれぞれの姿勢に応じて重み付けを付与したが、それぞれの姿勢が検知された継続時間に対して重み付けを付与する手法も採用し得る。例えば、１回として検知される犬座姿勢であっても、１分間で解かれる場合と１０分間継続する場合で、計数する回数を異ならせる。具体的には、例えば検知開始から１分の間は、検知時間（分）×１．０とし、その後３分までは、検知時間（分）×１．２とし、それ以降は、検知時間（分）×１．５などと設定する。このような重み付けを与えて総回数を計数することでも、より精度の高い体調不良予測が期待できる。この場合、姿勢による重み付けも併せて適用してもよい。 In the embodiment using the counting list 122'in FIG. 9, weighting is given according to each posture to be detected, but a method of giving weight to the duration in which each posture is detected can also be adopted. .. For example, even if the dog sitting posture is detected as one time, the number of times of counting is different depending on whether the posture is solved in 1 minute or continued for 10 minutes. Specifically, for example, the detection time (minutes) x 1.0 is set for 1 minute from the start of detection, the detection time (minutes) x 1.2 is set for up to 3 minutes thereafter, and the detection time (minutes) x 1.2 thereafter. Minutes) x 1.5, etc. By giving such weighting and counting the total number of times, more accurate prediction of poor physical condition can be expected. In this case, weighting by posture may also be applied.
　図１０は、更に他の実施例に係る豚飼育支援装置を採用した養豚環境の全体像を示す図である。図１と同様の要素については、同一符番を付すことによりその説明を省略する。 FIG. 10 is a diagram showing an overall picture of a pig farming environment in which a pig breeding support device according to another embodiment is adopted. The same elements as those in FIG. 1 are given the same numbering, and the description thereof will be omitted.
　図１０に示す実施例においては、飼育員は飼育員端末２２０を所持しておらず、代わりにそれぞれのペン３０１に隣接して告知灯２４０が一つずつ設置されている。それぞれの告知灯２４０は、無線ユニット２３０およびネットワーク２００を介してサーバ１００と接続される。サーバ１００は、例えば第５ペンにおいて計数した特定姿勢の回数が通知閾値を超えたら、第５ペンに隣接して設置された告知灯２４０へ超過通知に相当する通知信号を送信して当該告知灯２４０を点灯させる。このような告知灯２４０を利用すれば、飼育員は、飼育員端末２２０を所持していなくても向かうべきペン３０１を認識できる。飼育員は、告知灯２４０が点灯したペン３０１に収容された豚３０２を対象として、体調不良の兆候を示す豚を探せばよい。 In the embodiment shown in FIG. 10, the keeper does not have the keeper terminal 220, and instead, one notification light 240 is installed adjacent to each pen 301. Each notification light 240 is connected to the server 100 via the wireless unit 230 and the network 200. For example, when the number of times of the specific posture counted by the fifth pen exceeds the notification threshold value, the server 100 transmits a notification signal corresponding to the excess notification to the notification light 240 installed adjacent to the fifth pen, and the notification light. Turn on 240. By using such a notification light 240, the keeper can recognize the pen 301 to be directed even if he / she does not have the keeper terminal 220. The keeper may search for a pig showing signs of poor physical condition by targeting the pig 302 housed in the pen 301 in which the notification light 240 is turned on.
　図１１は、更に他の実施例に係る豚飼育支援装置の計数リスト１２２”を説明する図である。以上に説明した実施例おいては、１つのペン３０１に収容された複数の豚３０２は互いに区別して認識されないものであった。したがって、観察対象となる特定のペン３０１に収容されている豚３０２のいずれかが特定姿勢を取れば、１回の特定姿勢として検知した。例えば、通知閾値が５０回に設定されているときには、１頭の豚３０２が５０回を超える特定姿勢を取った場合も、１０頭の豚３０２がそれぞれ５～６回の特定姿勢を取った場合も、計数部１１３は超過通知を出力する。すなわち、飼育員は、いずれのペン３０１で規定値を超える特定姿勢が検知されたかがわかったとしても、特定の豚３０２が強く体調不良の兆候を示しているのか、ペン内の豚３０２が全体的に体調不良の傾向にあるのかを見極めることが必要である。 FIG. 11 is a diagram for explaining the counting list 122 ”of the pig breeding support device according to still another embodiment. In the above-described embodiment, the plurality of pigs 302 housed in one pen 301 are Therefore, if any of the pigs 302 housed in the specific pen 301 to be observed takes a specific posture, it is detected as one specific posture. For example, the notification threshold. When is set to 50 times, the counting unit regardless of whether one pig 302 takes a specific posture exceeding 50 times or 10 pigs 302 take a specific posture 5 to 6 times each. 113 outputs an excess notification. That is, even if the keeper knows which pen 301 has detected a specific posture exceeding the specified value, is the specific pig 302 strongly showing a sign of poor physical condition? It is necessary to determine whether the pig 302 in the pen tends to be in poor physical condition as a whole.
　一方で、それぞれのペン３０１内に収容された複数の豚３０２を互いに区別して認識する技術が知られるようになってきている。例えば、識別マーカをそれぞれの豚３０２に装着し、それをカメラユニット２１０が撮像して得た画像を解析することにより、個体識別を実現できる。あるいは、ペン３０１へ収容する時点でそれぞれの豚３０２を撮像して当該撮像画像に識別番号を対応させておき、その後に特定姿勢を検知した豚３０２がいずれの識別番号と対応付けられた撮像画像と適合するかを、学習モデルを用いて検知してもよい。 On the other hand, a technique for distinguishing and recognizing a plurality of pigs 302 housed in each pen 301 has become known. For example, individual identification can be realized by attaching an identification marker to each pig 302 and analyzing an image obtained by capturing the image with the camera unit 210. Alternatively, at the time of accommodating in the pen 301, each pig 302 is imaged and an identification number is associated with the captured image, and then the pig 302 that detects a specific posture is associated with any identification number. May be detected using a learning model.
　図１１に示す計数リスト１２２”は、それぞれのペン３０１内に収容された豚３０２の個体識別が可能な場合の計数リストである。それぞれのペン３０１には、例えば１０頭ずつの豚３０２が収容されており、それぞれの豚３０２は付与された識別番号によって区別される。そして、検知部１１２は、特定姿勢を検知すると共に、その特定姿勢を取った豚３０２の識別番号を特定する。計数部１１３は、検知部１１２からそれらの情報を受け取って、特定された識別番号に対応する豚の特定姿勢の回数を更新する。 The counting list 122 ”shown in FIG. 11 is a counting list when the individual pigs 302 housed in each pen 301 can be identified. For example, 10 pigs 302 are housed in each pen 301. Each pig 302 is distinguished by an assigned identification number. Then, the detection unit 112 detects a specific posture and identifies the identification number of the pig 302 that has taken the specific posture. The 113 receives the information from the detection unit 112 and updates the number of times of the specific posture of the pig corresponding to the specified identification number.
　なお、個体識別を行う場合には通知閾値を、個体識別を行わない場合に比べて小さな値に設定すれば、特定の豚３０２の体調不良をいち早く察知することができる。計数リスト１２２”では、通知閾値が２５回に設定されている。計数部１１３は、特定の豚３０２がこの通知閾値を超えた場合に、その識別番号の情報を付加した超過通知を飼育員端末２２０へ出力する。図１０を用いて説明した告知灯２４０が表示部を備えるのであれば、計数部１１３は、その特定の豚３０２が収容されたペンに隣接する告知灯２４０へ超過通知を出力し、その識別番号を当該告知灯２４０の表示部に表示してもよい。飼育員は、特定の豚３０２に関する情報も得ることができれば、ペン３０１に収容された複数の豚３０２の中から当該特定の豚３０２を容易に見つけ出すことができる。 If the notification threshold value is set to a smaller value when individual identification is performed than when individual identification is not performed, it is possible to quickly detect poor physical condition of a specific pig 302. In the counting list 122 ”, the notification threshold is set to 25 times. When the specific pig 302 exceeds this notification threshold, the keeper terminal sends an excess notification to which the information of the identification number is added. Output to 220. If the notification light 240 described with reference to FIG. 10 includes a display unit, the counting unit 113 outputs an excess notification to the notification light 240 adjacent to the pen in which the specific pig 302 is housed. Then, the identification number may be displayed on the display unit of the notification light 240. If the keeper can also obtain information on the specific pig 302, the keeper may select the pig 302 from among the plurality of pigs 302 housed in the pen 301. A specific pig 302 can be easily found.
　以上いくつかの実施例を通じて説明した本実施形態においては、それぞれのペン３０１に一つずつのカメラユニット２１０を設置したが、複数のペン３０１をまとめて俯瞰するカメラユニットを設置してもよい。その場合には、取得部１１１は、カメラユニットから取得した画像を各ペン３０１の境界に沿って分割し、分割したそれぞれの画像を順次検知部１１２へ引き渡せばよい。反対に、１つのペン３０１に対して複数のカメラユニット２１０を設置してもよい。例えば、ペン３０１内の豚３０２を側方から撮像するカメラユニット２１０を設置すれば、特定姿勢の誤検知を軽減できる。 In the present embodiment described through the above several examples, one camera unit 210 is installed for each pen 301, but a camera unit for overlooking a plurality of pens 301 may be installed. In that case, the acquisition unit 111 may divide the image acquired from the camera unit along the boundary of each pen 301, and sequentially deliver each of the divided images to the detection unit 112. On the contrary, a plurality of camera units 210 may be installed for one pen 301. For example, if a camera unit 210 that captures the pig 302 in the pen 301 from the side is installed, erroneous detection of a specific posture can be reduced.
　また、以上説明した本実施形態においては、複数のペン３０１が一つの豚舎に設けられている場合を想定したが、複数の豚舎に亘って設けられているペン３０１を観察対象にしてもよい。この場合に、豚舎の事情やそれぞれに収容されている豚の種類等によって、通知閾値や重み付けなどの設定値を異ならせてもよい。 Further, in the present embodiment described above, it is assumed that a plurality of pens 301 are provided in one piggery, but the pens 301 provided over the plurality of piggery may be an observation target. In this case, the setting values such as the notification threshold value and the weighting may be different depending on the circumstances of the piggery, the type of pigs housed in each, and the like.
　また、以上説明した本実施形態においては、基準観察時間が経過するごとにそれまでの記録をリセットして新たな基準観察時間では０から計数を始めたが、計数の手法はこれに限らない。例えば、基準観察時間（例えば２４時間）のうち、新たな単位時間（例えば１時間）の観察を開始する時点で最も古い単位時間分の観察結果を破棄するようにしてもよい。このように基準観察時間を時間の経過とともにシフトさせれば、現時点におけるそれぞれのペン３０１の状況をより正しく把握でき、体調不良の豚３０２が現れた場合もあまり遅延することなく超過通知を出力することができる。 Further, in the present embodiment described above, the record up to that point is reset every time the reference observation time elapses, and counting is started from 0 at the new reference observation time, but the counting method is not limited to this. For example, of the reference observation time (for example, 24 hours), the observation result for the oldest unit time at the time of starting the observation for a new unit time (for example, 1 hour) may be discarded. By shifting the reference observation time with the passage of time in this way, the current state of each pen 301 can be grasped more accurately, and even if an unwell pig 302 appears, an excess notification is output without much delay. be able to.
　また、以上説明した本実施形態において超過通知の出力先は、飼育員端末２２０であったり告知灯２４０であったりしたが、これらに限らない。計数部１１３は、サーバ１００に接続された表示モニタ１５０に、超過通知に関する情報を直接的に表示しても構わない。また、計数部１１３は、計数する特定姿勢の回数が通知閾値を超えた場合に超過通知を出力する場合に加え、あるいは超過通知を出力する代わりに、計数中の特定姿勢の回数を定常的に出力してもよい。例えば、それぞれのペン３０１において計数されている現時点の特定姿勢の回数が飼育員端末２２０に一覧表示されるようにしてもよい。 Further, in the present embodiment described above, the output destination of the excess notification is the keeper terminal 220 or the notification light 240, but the output destination is not limited to these. The counting unit 113 may directly display the information regarding the excess notification on the display monitor 150 connected to the server 100. Further, the counting unit 113 constantly counts the number of specific postures during counting in addition to the case where the excess notification is output when the number of times of the specific posture to be counted exceeds the notification threshold value, or instead of outputting the excess notification. It may be output. For example, the number of times of the current specific posture counted by each pen 301 may be displayed in a list on the keeper terminal 220.
　また、以上説明した本実施形態においては、特定姿勢として犬座姿勢、庇寝姿勢、背湾姿勢、パンティングの４つを対象としてその回数を計数したが、対象とする特定姿勢はこれに限らない。４つのうちのいずれかを選択してもよいし、体調不良の豚が示す他の特定姿勢をこれらに加えて、あるいはこれらに代えて設定してもよい。 Further, in the present embodiment described above, the number of times is counted for four specific postures: the dog sitting posture, the eaves posture, the back bay posture, and the punting, but the target specific posture is limited to this. do not have. One of the four may be selected, and other specific postures indicated by the unwell pig may be added to or replaced by these.
　また、以上説明した本実施形態においては、サーバ１００が豚飼育支援装置として機能する場合を説明したが、ハードウェア構成はこれに限らない。飼育員端末２２０として説明した携帯端末がサーバ１００と同様の処理を行えば、当該携帯端末が豚飼育支援装置として機能し得る。また、例えば、サーバ１００の処理の一部を飼育員端末２２０が担うように構成すれば、サーバ１００と飼育員端末２２０が連携するシステムが、豚飼育支援装置となり得る。 Further, in the present embodiment described above, the case where the server 100 functions as a pig breeding support device has been described, but the hardware configuration is not limited to this. If the mobile terminal described as the keeper terminal 220 performs the same processing as the server 100, the mobile terminal can function as a pig breeding support device. Further, for example, if the keeper terminal 220 is configured to take part in the processing of the server 100, the system in which the server 100 and the keeper terminal 220 cooperate with each other can be a pig breeding support device.
１００…サーバ、１１０…演算部、１１１…取得部、１１２…検知部、１１３…計数部、１２０…記憶部、１２１…検知用ニューラルネットワーク、１２２…計数リスト、１３０…通信ユニット、１５０…表示モニタ、１６０…入力デバイス、２００…ネットワーク、２１０…カメラユニット、２２０…飼育員端末、２３０…無線ユニット、２４０…告知灯、３０１…ペン、３０２…豚 100 ... server, 110 ... arithmetic unit, 111 ... acquisition unit, 112 ... detection unit, 113 ... counting unit, 120 ... storage unit, 121 ... detection neural network, 122 ... counting list, 130 ... communication unit, 150 ... display monitor , 160 ... Input device, 200 ... Network, 210 ... Camera unit, 220 ... Keeper terminal, 230 ... Wireless unit, 240 ... Notification light, 301 ... Pen, 302 ... Pig

Claims (13)

  1. 　それぞれにおいて豚が集団飼育されている複数のペンに向けられて設置されたカメラによって撮像された画像の画像データを取得する取得部と、
    　前記画像データの画像に基づいて前記豚の特定姿勢を検知する検知部と、
    　設定された観察時間の間に前記複数のペンのそれぞれにおいて検知された前記特定姿勢の回数を計数する計数部と
    を備える豚飼育支援装置。     In each case, there is an acquisition unit that acquires image data of images taken by cameras installed toward multiple pens in which pigs are group-reared.
    A detection unit that detects the specific posture of the pig based on the image of the image data,
    A pig breeding support device including a counting unit that counts the number of times of the specific posture detected by each of the plurality of pens during a set observation time.
  2. 　前記計数部は、前記複数のペンのうちいずれかのペンにおいて前記回数が設定された閾値を超えた場合に、当該ペンに関するペン情報を含む超過通知を出力する請求項１に記載の豚飼育支援装置。 The pig breeding support according to claim 1, wherein the counting unit outputs an excess notification including pen information regarding the pen when the number of times exceeds a set threshold value in any of the plurality of pens. Device.
  3. 　前記計数部は、集団飼育されているそれぞれの前記豚を識別して前記特定姿勢の回数を計数し、前記超過通知を出力する場合には前記特定姿勢を示した前記豚に関する識別情報を前記超過通知に付加する請求項２に記載の豚飼育支援装置。 The counting unit identifies each of the pigs in a group and counts the number of times of the specific posture, and when outputting the excess notification, the counting unit exceeds the identification information about the pig showing the specific posture. The pig breeding support device according to claim 2, which is added to the notification.
  4. 　前記検知部は、前記特定姿勢として前脚を立てた状態で着座する犬座姿勢を検知する請求項１から３のいずれか１項に記載の豚飼育支援装置。 The pig breeding support device according to any one of claims 1 to 3, wherein the detection unit detects a dog sitting posture in which the front leg is upright as the specific posture.
  5. 　前記検知部は、前記特定姿勢として伏臥した胴部の下に前脚を収める庇寝姿勢を検知する請求項１から４のいずれか１項に記載の豚飼育支援装置。 The pig breeding support device according to any one of claims 1 to 4, wherein the detection unit detects a lying posture in which the front legs are housed under the prone body as the specific posture.
  6. 　前記検知部は、前記特定姿勢として背中を丸めて前屈みとなる背湾姿勢を検知する請求項１から５のいずれか１項に記載の豚飼育支援装置。 The pig breeding support device according to any one of claims 1 to 5, wherein the detection unit detects a back bay posture in which the back is curled up and bends forward as the specific posture.
  7. 　前記検知部は、前記特定姿勢として横臥位で腹部を波打たせたパンティングを検知する請求項１から６のいずれか１項に記載の豚飼育支援装置。 The pig breeding support device according to any one of claims 1 to 6, wherein the detection unit detects wheezing of the abdomen in a lying position as the specific posture.
  8. 　前記検知部は、連続する複数の前記画像に基づいて前記特定姿勢を検知する請求項１から７のいずれか１項に記載の豚飼育支援装置。 The pig breeding support device according to any one of claims 1 to 7, wherein the detection unit detects the specific posture based on a plurality of continuous images.
  9. 　前記検知部は、前記特定姿勢を取る豚が写った教師画像によって学習された学習モデルを用いて前記特定姿勢を検知する請求項１から８のいずれか１項に記載の豚飼育支援装置。 The pig breeding support device according to any one of claims 1 to 8, wherein the detection unit detects the specific posture by using a learning model learned from a teacher image of a pig taking the specific posture.
  10. 　前記計数部は、前記検知部が前記特定姿勢として複数種類の姿勢を対象として検知する場合には、検知した姿勢に応じて重み付けを付与して計数する請求項１から９のいずれか１項に記載の豚飼育支援装置。 According to any one of claims 1 to 9, when the detecting unit detects a plurality of types of postures as the specific postures, the counting unit assigns weights according to the detected postures and counts the postures. The pig breeding support device described.
  11. 　前記計数部は、前記検知部が前記特定姿勢を継続して検知した場合には、検知した継続時間に応じて重み付けを付与して計数する請求項１から１０のいずれか１項に記載の豚飼育支援装置。 The pig according to any one of claims 1 to 10, wherein when the detection unit continuously detects the specific posture, the counting unit weights and counts the pig according to the detected duration. Breeding support device.
  12. 　それぞれにおいて豚が集団飼育されている複数のペンに向けられて設置されたカメラによって撮像された画像の画像データを取得する取得ステップと、
    　前記画像データの画像に基づいて前記豚の特定姿勢を検知する検知ステップと、
    　設定された観察時間の間に前記複数のペンのそれぞれにおいて検知された前記特定姿勢の回数を計数する計数ステップと
    を有する豚飼育支援方法。     In each case, the acquisition step of acquiring the image data of the image taken by the camera installed toward the pens in which the pigs are group-reared, and
    A detection step for detecting the specific posture of the pig based on the image of the image data, and
    A pig breeding support method having a counting step for counting the number of times of the specific posture detected by each of the plurality of pens during a set observation time.
  13. 　それぞれにおいて豚が集団飼育されている複数のペンに向けられて設置されたカメラによって撮像された画像の画像データを取得する取得ステップと、
    　前記画像データの画像に基づいて前記豚の特定姿勢を検知する検知ステップと、
    　設定された観察時間の間に前記複数のペンのそれぞれにおいて検知された前記特定姿勢の回数を計数する計数ステップと
    をコンピュータに実行させる豚飼育支援プログラム。     In each case, the acquisition step of acquiring the image data of the image taken by the camera installed toward the pens in which the pigs are group-reared, and
    A detection step for detecting the specific posture of the pig based on the image of the image data, and
    A pig breeding support program that causes a computer to perform a counting step that counts the number of times of the specific posture detected by each of the plurality of pens during a set observation time.
PCT/JP2021/047913 2021-01-18 2021-12-23 Pig rearing assistance apparatus, pig rearing assistance method, and pig rearing assistance program WO2022153829A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2022575504A JPWO2022153829A1 (en) 2021-01-18 2021-12-23

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2021005449 2021-01-18
JP2021-005449 2021-01-18

Publications (1)

Publication Number Publication Date
WO2022153829A1 true WO2022153829A1 (en) 2022-07-21

Family

ID=82448445

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2021/047913 WO2022153829A1 (en) 2021-01-18 2021-12-23 Pig rearing assistance apparatus, pig rearing assistance method, and pig rearing assistance program

Country Status (2)

Country Link
JP (1) JPWO2022153829A1 (en)
WO (1) WO2022153829A1 (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120326862A1 (en) * 2011-06-22 2012-12-27 Hana Micron America Inc. Early Alert System and Method for Livestock Disease Detection
WO2016121096A1 (en) * 2015-01-30 2016-08-04 株式会社コムテック Method and device for detecting abnormal state of pig
JP2019024482A (en) * 2017-07-31 2019-02-21 株式会社コンピューター総合研究所 Information processing system, information processing device, and program
WO2019058752A1 (en) * 2017-09-22 2019-03-28 パナソニックＩｐマネジメント株式会社 Livestock information management system, livestock barn, livestock information management program, and livestock information management method
JP2020014421A (en) * 2018-07-26 2020-01-30 日本ユニシス株式会社 Livestock birth prediction system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120326862A1 (en) * 2011-06-22 2012-12-27 Hana Micron America Inc. Early Alert System and Method for Livestock Disease Detection
WO2016121096A1 (en) * 2015-01-30 2016-08-04 株式会社コムテック Method and device for detecting abnormal state of pig
JP2019024482A (en) * 2017-07-31 2019-02-21 株式会社コンピューター総合研究所 Information processing system, information processing device, and program
WO2019058752A1 (en) * 2017-09-22 2019-03-28 パナソニックＩｐマネジメント株式会社 Livestock information management system, livestock barn, livestock information management program, and livestock information management method
JP2020014421A (en) * 2018-07-26 2020-01-30 日本ユニシス株式会社 Livestock birth prediction system

Also Published As

Publication number Publication date
JPWO2022153829A1 (en) 2022-07-21

Similar Documents

Publication Publication Date Title
Berckmans Automatic on-line monitoring of animals by precision livestock farming
Grosso et al. On-farm Qualitative Behaviour Assessment of dairy goats in different housing conditions
US9538729B2 (en) Cattle monitoring for illness
Cornou et al. Modelling and monitoring sows’ activity types in farrowing house using acceleration data
JP2022514115A (en) Livestock surveillance
Nasirahmadi et al. Using automated image analysis in pig behavioural research: Assessment of the influence of enrichment substrate provision on lying behaviour
Berckmans et al. Animal sound… talks! Real-time sound analysis for health monitoring in livestock
Papadakis et al. Sub-second analysis of fish behavior using a novel computer-vision system
Meyer et al. Development and validation of broiler welfare assessment methods for research and on-farm audits
Berckmans 1.2. Smart farming for Europe: value creation through precision livestock farming
KR20200071597A (en) Prediction method and the apparatus for onset time of sow farrowing by image analysis
US20230413786A1 (en) Pig rearing support apparatus, pig rearing support method, and non-volatile storage medium storing pig rearing support program
Mittek et al. Health monitoring of group-housed pigs using depth-enabled multi-object tracking
WO2022153829A1 (en) Pig rearing assistance apparatus, pig rearing assistance method, and pig rearing assistance program
DeBoer et al. Does the presence of a human affect the preference of enrichment items in young, isolated pigs?
JP7162749B2 (en) Estrus determination device for sows, estrus determination method for sows, and estrus determination program for sows
Galeana et al. Mother-young spatial association and its relation with proximity to a fence separating ewes and lambs during enforced weaning in hair sheep (Ovis aries)
Ojukwu et al. Development of a computer vision system to detect inactivity in group-housed pigs
JP7504036B2 (en) Pig breeding support device, pig breeding support method, and pig breeding support program
Bello et al. A framework for real-time cattle monitoring using multimedia networks
JP7445612B2 (en) Pig breeding support device, pig breeding support method, and pig breeding support program
Ismayilova The use of image labelling to identify pig behaviours for the development of a real-time monitoring and control tool
JP2023031517A (en) Information processing system, information processing method and program
JP7450561B2 (en) Pig breeding support device, pig breeding support method, and pig breeding support program
Siegford et al. Practical considerations for the use of precision livestock farming to improve animal welfare

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21919716

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2022575504

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21919716

Country of ref document: EP

Kind code of ref document: A1