JP7211595B2 - Abnormal behavior detection device, abnormal behavior detection program, and abnormal behavior detection method - Google Patents

Description

TECHNICAL FIELD The present invention relates to an abnormal behavior detection device, an abnormal behavior detection program, and an abnormal behavior detection method suitable for detecting abnormal behavior such as so-called loss of appetite and estrous behavior based on the movement trajectory of an animal such as a dairy cow. .

Conventionally, as a technology for detecting abnormalities in dairy cows, for example, Japanese Patent Application Laid-Open No. 2019-122368 discloses an activity state sensor module attached to a cow and a management device connected via a communication network. A management device calculates a behavioral index, a static index, and a rumination index corresponding to the specified activity state of the cow based on the activity data of the cow received from the activity state sensor module, and based on these indexes Japanese Patent Laid-Open No. 2002-100003 discloses a health condition management system that determines whether or not there is an abnormality in the health condition of a cow.

JP 2019-122368 A

However, in the health condition management system described in Patent Document 1, abnormal health conditions (disease, poor physical condition, etc.) are simply detected based on the cow's activity state (stillness, walking, feeding, etc.) estimated by an acceleration sensor or the like. ). For this reason, weak cows feeding at the feeding area in the barn are pushed out by stronger cows (so-called "eaten"), and behaviors such as rubbing against adjacent dairy cows in the aisle of the barn ( There is a problem that it is not possible to detect special abnormal behaviors related to living areas related to the lives of dairy cows, such as estrus behavior).

SUMMARY OF THE INVENTION The present invention has been made to solve such problems. An object of the present invention is to provide a behavior detection program and an abnormal behavior detection method.

In order to solve the problem of detecting abnormal behavior such as eating and estrous behavior based on the movement trajectory of the animal, the abnormal behavior detection device according to the present invention provides movement trajectory data indicating the movement trajectory of each individual animal. and living area data defining position information of living areas related to the life of the animal, such as feeding areas, passages, beds, etc., for creating detection data for detecting abnormal behavior of the animal. A data creation unit and an abnormal behavior detection unit that detects abnormal behavior of the animal based on the detection data.

Further, as one aspect of the present invention, in order to solve the problem of detecting eating loss based on the elapsed time and movement distance in the living area where the animal stayed, the detection data creation unit includes: creates detection data including the living area where the person stayed, the start date and time of stay, the end date and time of stay in the living area, and the distance traveled in the living area, and the abnormal behavior detection unit detects the living area If an individual that moves from one of the feeding areas to another feeding area via the passage area moves a distance that exceeds the specified value within the elapsed time that is less than the specified value, even if it is detected as an abnormal behavior good.

Furthermore, as one aspect of the present invention, in order to solve the problem of detecting an abnormal behavior based on a trajectory diagram of an animal and a convolutional neural network, the detection data creation unit is configured to: Data for detection including the moved trajectory diagram is created, and the abnormal behavior detection unit learns the trajectory diagram when abnormal behavior does not occur and the trajectory diagram when abnormal behavior occurs in the convolutional neural network. Abnormal behavior may be detected by inputting the trajectory diagram into a trained neural network that has undergone supervised learning as data.

Further, as one aspect of the present invention, in order to solve the problem of detecting an abnormal behavior based on a trajectory diagram of an animal and a convolutional autoencoder, the detection data generation unit is configured to: Data for detection including a moving trajectory diagram is created, and the abnormal behavior detection unit uses the trajectory diagram when no abnormal behavior occurs in the convolutional autoencoder as learning data for the learned auto-encoder that has undergone unsupervised learning. Abnormal behavior may be detected by inputting the trajectory diagram to an encoder.

Furthermore, as one aspect of the present invention, in order to solve the problem of detecting the estrus behavior based on the time period and range in which the animal stayed in the passage area, the detection data creation unit The detection data including the period of time spent in the area and the range in which the animal stayed in the passage area is created, and the abnormal behavior detection unit detects the time spent in the passage area in the detection data of different individuals. When both the band and the range overlap, and when the distance within the predetermined value occurs more than a predetermined number of times on the same day, it may be detected as an estrous behavior.

An abnormal behavior detection program according to the present invention solves the problem of detecting abnormal behavior such as eating and estrous behavior based on the movement trajectory of an animal. and living area data defining position information of living areas related to the life of the animal, such as feeding areas, passages, beds, etc., for creating detection data for detecting abnormal behavior of the animal. The computer is caused to function as a data creation unit and an abnormal behavior detection unit that detects abnormal behavior of the animal based on the detection data.

The abnormal behavior detection method according to the present invention is based on the movement trajectory of an animal. and living area data defining position information of living areas related to the life of the animal, such as feeding areas, passages, beds, etc., for creating detection data for detecting abnormal behavior of the animal. and an abnormal behavior detection step of detecting abnormal behavior of the animal based on the detection data.

According to the present invention, it is possible to detect abnormal behavior such as eating and estrous behavior based on the movement trajectory of the animal.

1 is a diagram showing an embodiment of an abnormal behavior detection device and an abnormal behavior detection program according to the present invention; FIG. It is a figure which shows the living area defined in the 1st Embodiment of this invention. It is a figure which shows an example of the living area data of the 1st Embodiment. It is a figure which shows an example of the movement locus|trajectory data of the 1st Embodiment of this invention. It is a figure which shows an example of the data for detection of the 1st Embodiment of this invention. 6 is a flowchart showing detection data creation processing according to the first embodiment; 4 is a flowchart showing abnormal behavior detection processing according to the first embodiment; FIG. 10 is a diagram showing how the convolutional neural network according to the second embodiment is subjected to supervised learning of learning data; FIG. 12 is a diagram showing how the trained convolutional neural network of the second embodiment determines whether or not there is an abnormal behavior in the trajectory diagram. 9 is a flow chart showing detection data creation processing and abnormal behavior detection processing according to the second embodiment. FIG. 10 is a diagram showing how a convolutional autoencoder learns training data without supervision; FIG. 10 is a diagram showing how a trained convolutional autoencoder determines whether or not there is abnormal behavior in a trajectory diagram. FIG. 12 is a diagram showing the relationship between detection data and movement locus data according to the third embodiment; FIG. 14 is a flowchart showing abnormal behavior detection processing according to the third embodiment; FIG.

Hereinafter, embodiments of an abnormal behavior detection device, an abnormal behavior detection program, and an abnormal behavior detection method according to the present invention will be described with reference to the drawings.

The abnormal behavior detection device 1 according to the present invention is for detecting abnormal behavior based on the locus of movement of an animal and notifying the user terminal 11 of the detection information. In each of the following embodiments, the animal to be detected is assumed to be a dairy cow in a barn, but the present invention is not limited to this, and can be applied to any animal whose abnormal behavior can be detected based on its movement trajectory. The invention applies.

The abnormal behavior detection device 1 of the first embodiment is configured by a computer such as a personal computer, a smartphone, or a tablet, and as shown in FIG. 4. Further, the abnormal behavior detection device 1 is connected to the user terminal 11 so as to be able to communicate wirelessly, and notifies the user terminal 11 when abnormal behavior of the animal is detected. Each component will be described in detail below.

The user terminal 11 is configured by a computer such as a personal computer, a smart phone, or a tablet, and is a terminal owned or carried by a user. In the first embodiment, the user terminal 11 receives detection information from the abnormal behavior detection device 1 and notifies the user of the detection information by voice, display, vibration, or the like.

The communication means 2 implements a communication function in the abnormal behavior detection device 1 . In the first embodiment, the communication means 2 is composed of a wireless module compatible with a wireless LAN standard such as Wi-Fi (registered trademark) or a carrier network standard such as a fifth generation mobile communication network (5G). . Then, it communicates wirelessly with the user terminal 11 via an access point or the like (not shown).

The storage means 3 stores various data and functions as a working area when the arithmetic processing means 4 performs arithmetic processing. In the first embodiment, the storage means 3 includes a hard disk, ROM (Read Only Memory), RAM (Random Access Memory), flash memory, etc. As shown in FIG. It has a living area data storage unit 32 , a movement track data storage unit 33 , a detection data storage unit 34 and a user data storage unit 35 . Each component will be described in more detail below.

The program storage unit 31 is installed with the abnormal behavior detection program 1a of the first embodiment. The arithmetic processing unit 4 executes the abnormal behavior detection program 1a to cause the computer as the abnormal behavior detection device 1 to function as each component described later.

The usage form of the abnormal behavior detection program 1a is not limited to the configuration described above. For example, the abnormal behavior detection program 1a may be stored in a non-temporary computer-readable recording medium such as a CD-ROM or DVD-ROM, and may be directly read out from the recording medium and executed. Also, it may be used from an external server or the like by a cloud computing method or an ASP (Application Service Provider) method.

The living area data storage unit 32 defines the positional information of the living area related to the life of the dairy cow (animal), such as feeding grounds, passageways, beds, and the like. In the first embodiment, as shown in FIG. 2, a feeding area, passage area, bed area, and water fountain area are set in the barn as living areas. In this case, as living area data, as shown in FIG. 3, an area type and position information are defined in association with an area ID that identifies each area.

In the first embodiment, the positional information of each section has the upper left point in the cowshed in FIG. Defined by length and Y-direction length. However, the present invention is not limited to this, and any information that can specify the range in which each area exists may be used.

The trajectory data storage unit 33 stores trajectory data indicating the trajectory of each dairy cow (animal). In the first embodiment, as movement trajectory data, as shown in FIG. 4, the date and time and the position at which the dairy cow (animal) was present are associated with an individual ID that identifies each individual cow (animal). Information (X coordinate and Y coordinate) is created one record at a time at predetermined time intervals. Further, the moving track data storage unit 33 stores the moving track data for all dairy cows (animals) to be monitored. The position information of each animal may be obtained from a sensor or the like, or may be obtained from images of the inside of the barn taken at predetermined time intervals.

The detection data storage unit 34 stores detection data for detecting abnormal behavior of dairy cows (animals). In the first embodiment, the detection data is created by the detection data creation unit 41, which will be described later, for the purpose of reducing the amount of data held. Specifically, if a plurality of continuous records (coordinate points) in the movement trajectory data are within the same area and the movement distance (distance between coordinate points) is less than or equal to a predetermined value, it is regarded as not having moved. The detection data are collected into one record.

In the first embodiment, detection data as shown in FIG. 5 is created based on the living area data of FIG. 3 and the movement track data of FIG. Specifically, in association with the individual ID, the stay start date and time, the stay end date and time, and the position information (X coordinate and Y coordinate) where the dairy cow (animal) existed in the time period from the stay start date and time to the stay end date and time , the movement distance, which is the distance between the position of the cow at the start date and time of stay and the position of the cow at the end date and time of stay, and the zone ID of the zone where the cow (animal) was present.

The user data storage unit 35 stores user data necessary for transmitting detection information to the user. In the first embodiment, as user data, an e-mail address for receiving detection information on the user terminal 11, account information of SNS (Social Networking Service), and the like are registered.

Next, the arithmetic processing means 4 is composed of a CPU (Central Processing Unit) or the like, and by executing the abnormal behavior detection program 1a installed in the storage means 3, as shown in FIG. It functions as a creation unit 41 , an abnormal behavior detection unit 42 , and a detection information transmission unit 43 .

The detection data creation unit 41 creates detection data for detecting abnormal behavior of dairy cows (animals). In the first embodiment, the detection data creation unit 41 determines the living area where the dairy cow (animal) stayed, the start date and time of stay in the living area, and Detection data including the end date and time and the moving distance, which is the distance between the position of the cow at the start date and time of the stay and the position of the cow at the end date and time of the stay, is created.

Specifically, as shown in FIG. By referring to the living area data in the record, it is determined in which living area the point indicated by the record is included. As a result, if the determined living area is different from the previous point (including the case where the previous point does not exist), or even if it is the same area, the moving distance from the previous point is a predetermined value. , the detection data creation unit 41 adds a new record to the detection data storage unit 34 and stores it as detection data.

In the new record added as the detection data, the dates and times of the movement trajectory data are written as the stay start date and time and the stay end date and time. As the position information (X coordinate and Y coordinate), the position information (X coordinate and Y coordinate) of the trajectory data is written. Furthermore, the moving distance is written with the distance between the point of the previous record stored in the detection data and the acquired point. Also, in the area ID, an area ID corresponding to the living area determined by the detection data creation unit 41 is written.

On the other hand, if the determined living area is the same area as the previous point (record) and the moving distance from the previous point is equal to or less than a predetermined value, a new record is stored in the detection data storage unit 34. Instead of adding, the stay end date and time of the registered record already stored in the detection data storage unit 34 is rewritten and updated with the date and time of the movement trajectory data. The detection data creation unit 41 repeatedly executes the above processing for all records of the movement trajectory data.

The abnormal behavior detection unit 42 detects abnormal behavior of dairy cows (animals) based on detection data. In the first embodiment, as shown in FIG. 1, the abnormal behavior detection unit 42 acquires the detection data stored in the detection data storage unit 34 record by record, and satisfies the condition corresponding to abnormal behavior. Determine whether or not Then, when abnormal behavior is detected, the abnormal behavior detection unit 42 outputs detection information including the individual ID included in the detection data to the detection information transmission unit 43 .

In the first embodiment, as an abnormal behavior, "eating" which greatly affects the physical condition of dairy cows is set as a detection target. In general, strong dairy cows that are not overwhelmed tend to eat without moving too much from multiple feeding grounds. On the other hand, weaker dairy cows often move from one feeding station to another.

Therefore, in the first embodiment, as a detection condition for detecting eating and losing as abnormal behavior, an individual that has moved from a feeding ground area to another feeding ground area via a passage area in the living area has a predetermined If the movement distance exceeds a predetermined value within the elapsed time equal to or less than the value, it is detected as abnormal behavior. It should be noted that the detection conditions are not limited to the above, and it is possible to appropriately set conditions corresponding to various abnormal behaviors.

The detection information transmission unit 43 transmits the detection information to the user terminal 11 and notifies the user terminal 11 when abnormal behavior of the dairy cow (animal) is detected. In the first embodiment, the detection information transmission unit 43 acquires user data registered in the user data storage unit 35 as shown in FIG. , the detection information including the individual ID of the individual exhibiting the abnormal behavior is transmitted via the communication means 2 .

Next, the operation of the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the first embodiment will be described.

When abnormal behavior of a dairy cow (animal) is detected using the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the first embodiment, the detection data generation process is performed by the detection data generation unit 41. After execution (detection data creation step), abnormal behavior detection processing is executed by the abnormal behavior detection unit 42 (abnormal behavior detection step). Specific processing contents of each step will be described below.

In the detection data creation step, as shown in FIG. 6, the detection data creation unit 41 first acquires one record of movement trajectory data from the movement trajectory data storage unit 33 (step S1). Next, the detection data creation unit 41 refers to the living area data in the living area data storage unit 32, and determines which living area includes the point indicated by the record acquired in step S1 (step S2).

If the living area determined in step S2 is the same area as the point indicated by the previous record, and the moving distance from the point is equal to or less than a predetermined value (step S3: YES), detection data is created. The unit 41 rewrites and updates the movement end date and time of the registered record stored in the detection data storage unit 34 to the date and time of the movement trajectory data acquired in step S1 (step S4).

On the other hand, if the living area determined in step S2 is different from the point indicated by the previous record, or even if it is the same area, if the moving distance from the previous point exceeds a predetermined value (step S3: NO), the detection data creation unit 41 calculates the movement distance from the previous record (step S5), adds a new record to the detection data storage unit 34, and saves it as detection data. (step S6).

After the processing from step S1 to step S6 described above is repeatedly executed for all records of the movement trajectory data (step S7: YES), this processing ends. As a result, among a plurality of continuous records in the movement trajectory data, a group of records with a short movement distance within the same area is combined into one record and compressed as detection data. Therefore, the amount of data to be held is reduced, and there is no need to increase the capacity of the storage means 3 .

In the abnormal behavior detection step, as shown in FIG. 7, first, the abnormal behavior detection unit 42 initializes the parameter L indicating the movement distance (step S11), and then stores the detection data from the detection data storage unit 34 by 1. A record is obtained (step S12). If the zone ID included in the detection data is not a passage zone (step S13: NO), the process proceeds to step S15. On the other hand, if the area ID included in the detection data indicates a passage area (step S13: YES), the parameter L is updated by adding the movement distance included in the detection data (step S14).

Next, the abnormal behavior detection unit 42 determines whether or not the condition that the area ID included in the detection data is the feeding area area and the previous record is the passage area is satisfied (step S15). ), if the condition is not satisfied (step S15: NO), the process proceeds to step S19. On the other hand, if the above conditions are satisfied (step S15: YES), the abnormal behavior detection unit 42 satisfies the elapsed time by subtracting the starting date and time Ts, which is the starting point of the elapsed time, from the stay start date and time included in the detection data. The parameter T shown is set and updated (step S16).

Then, when the elapsed time T is equal to or less than the predetermined threshold value _T0 and the movement distance exceeds the predetermined threshold value _L0 (step S17: YES), the abnormal behavior detection unit 42 detects the individual ID included in the detection data. is output to the detection information transmission unit 43 (step S18). On the other hand, if the elapsed time T exceeds the predetermined threshold value _T0 or the movement distance is equal to or less than the predetermined threshold value _L0 (step S17: NO), the abnormal behavior detection unit 42 does not detect abnormal behavior and performs step S19. proceed to

After that, if the area ID included in the detection data is not the feeding area area (step S19: NO), the abnormal behavior detection unit 42 proceeds to step S21. On the other hand, if the area ID included in the detection data is the feeding area (step S19: YES), the start date and time Ts is set to the stay end date and time of the detection data and updated, and the movement distance is initialized (step S20). After the processing from step S11 to step S20 described above is repeatedly executed for all records of the detection data (step S21: YES), this processing ends.

With the abnormal behavior detection process described above, an individual that has moved from a feeding ground area to another feeding ground area via an aisle area and that has moved beyond a predetermined distance within a predetermined period of time will not be eaten. Because it is thought that it moved from the original feeding ground to another feeding ground, it is detected as an abnormal behavior. Then, the detection information is transmitted to the user terminal 11 via the communication means 2 by the detection information transmission unit 43 . Thereby, the abnormal behavior of the dairy cow (animal) is sequentially notified to the user.

Further, the detection information received by the user terminal 11 includes the individual ID of the individual that exhibits the abnormal behavior. Therefore, it is also possible to specify the current position of the individual corresponding to the individual ID by referring to the movement trajectory data corresponding to the individual ID. In actual system operation, records are added to the movement trajectory data as time passes, so the abnormal behavior detection process is executed sequentially at predetermined time intervals.

According to the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the first embodiment as described above, the following effects are obtained.
1. Based on the movement trajectories of animals, special abnormal behaviors related to living areas can be detected with high accuracy.
2. Eating or losing can be detected based on the elapsed time or distance traveled within the living area where the animal stayed.
3. The current location of an individual exhibiting abnormal behavior can be identified.
4. Even when a large number of dairy cows are raised in a large barn such as a free stall barn or a free barn barn, it is possible to identify all dairy cows individually and monitor the occurrence of abnormal behavior over a wide area. can.

Next, a second embodiment of the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method according to the present invention will be described. In addition, the same code|symbol is attached|subjected about the same or corresponding structure as 1st Embodiment mentioned above among the structures of the 2nd Embodiment, and description for the second time is abbreviate|omitted.

The features of the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the second embodiment are as an algorithm of the abnormal behavior detection unit 42 in the first embodiment and as a deep learning network model. It uses a known convolutional neural network (CNN).

In the second embodiment, the detection data creation unit 41 creates detection data including a trajectory diagram (image) of the animal's movement within a predetermined time based on the movement trajectory data and the living area data. . In the trajectory diagram, in addition to the polygonal line based on the coordinate point sequence of the movement trajectory data, the boundary line of each living area is drawn. Further, in order to highlight the characteristics of the movement trajectory, each living area may be color-coded by area type, or the coordinate points may be color-coded according to which area type the living area is included in. In the second embodiment, the trajectory diagram in the feeding ground area is used as detection data, but the data is not limited to this, and it is possible to determine how the animal moved within the living area or between living areas. It is sufficient if it is a trajectory diagram representing

Further, in the second embodiment, as shown in FIG. 8, a trajectory diagram (normal data) when abnormal behavior does not occur and a trajectory diagram (abnormal data) when abnormal behavior (e.g., eating and losing) occurs. data) in large quantities. Then, a trained neural network obtained by supervised learning of a convolutional neural network using these as learning data is set as an algorithm of the abnormal behavior detection unit 42 .

As for the trajectory diagram when no abnormal behavior has occurred, for example, a trajectory diagram near the feeding ground generated from the movement trajectory data of strong individuals who are not overwhelmed is taken as normal data. On the other hand, as a trajectory diagram when abnormal behavior occurs, a trajectory diagram near the feeding ground generated from movement trajectory data of a weak individual that is known to actually eat is regarded as abnormal data.

With the above configuration, as shown in FIG. 9, the abnormal behavior detection unit 42 detects abnormal behavior by inputting the trajectory diagram created by the detection data creation unit 41 into the above-described learned neural network. do. That is, the abnormal behavior detection unit 42 determines that the behavior is abnormal only when a trajectory diagram when abnormal behavior occurs is input, and outputs detection information.

Next, functions of the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the second embodiment will be described.

In the second embodiment, first, as shown in FIG. 10, the detection data creation unit 41 acquires one record of movement trajectory data from the movement trajectory data storage unit 33 (step S31). Next, the detection data creation unit 41 refers to the living area data in the living area data storage unit 32, and determines which living area includes the point indicated by the record acquired in step S31 (step S32).

If the living area determined in step S32 is the feeding area (step S33: YES), the detection data creation unit 41 creates a trajectory diagram (image) of movement within the most recent predetermined time (step S34). ), the trajectory diagram is input to the learned neural network (step S35). As a result, the abnormal behavior of the animal that can be read from the trajectory diagram is detected by the abnormal behavior detection unit 42 . As a result, when it is detected that an abnormal behavior (eating loss) is occurring (step S36: YES), detection information including the individual ID included in the detection data is output to the detection information transmitting unit 43 (step S37). ).

After the above processing is repeatedly executed for all records of the movement trajectory data (step S38: YES), this processing is terminated. If the living area determined in step S32 is not the feeding area (step S33: NO) or if no abnormal behavior is detected (step S36: NO), the process proceeds directly to step S38.

According to the second embodiment as described above, the same effects as those of the above-described first embodiment are obtained using a machine learning algorithm.

In the above-described second embodiment, a convolutional neural network is used as a deep learning network model, but the present invention is not limited to this, and a convolutional autoencoder (CAE) is used. may be used. When using a convolutional autoencoder, as shown in FIG. 11, only the trajectory diagram when abnormal behavior (eating and losing) does not occur is used as learning data, and the error (average Unsupervised learning is performed so that the squared error) is minimized.

As a result, in the learned autoencoder, as shown in FIG. 12, the error between the restored image and the original image becomes large only when the trajectory diagram when abnormal behavior (eating and losing) occurs is input. . Therefore, by setting an appropriate threshold value, it is determined that the trajectory diagram (image) when abnormal behavior (eating and losing) is occurring is abnormal. Note that the convolutional autoencoder performs machine learning using only trajectory diagrams (normal data) in which abnormal behavior does not occur. For this reason, desired learning data can be easily prepared by, for example, generating a large number of trajectory diagrams in the vicinity of the feeding ground from movement trajectory data of strong individuals that do not lose out.

Next, a third embodiment of the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method according to the present invention will be described. In addition, among the configurations of the third embodiment, configurations that are the same as or correspond to those of the above-described embodiments are denoted by the same reference numerals, and descriptions thereof will be omitted.

The abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the third embodiment are characterized in that estrous behavior is detected as abnormal behavior of dairy cows (animals).

In the third embodiment, the detection data creation unit 41 generates a time zone (stay start date and time and stay end date and time) and the extent of the animal's stay within the corridor area. The range of stay is the minimum coordinates (Xmin, Ymin) and the maximum coordinates (Xmax, Ymax) of the coordinate point sequence indicated by the movement trajectory data. Further, in the third embodiment, the detection data also includes the number of coordinate points.

In general, dairy cows in estrus often take actions such as snuggling their bodies against adjacent dairy cows in the aisle of the barn. Therefore, in the third embodiment, when a certain level of similarity or more is found in the detection data of individuals different from each other, the abnormal behavior detection unit 42 specifically detects the time zone and range of stay in the passage area. If there is an overlap between the two and the two individuals are within a predetermined distance on the same date and time more than a predetermined number of times, it is detected as an estrous behavior. The distance between the two individuals is calculated by the distance between the coordinate points in the movement trajectory data of each individual.

Next, functions of the abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method of the third embodiment will be described.

In the third embodiment, first, as shown in FIG. 14, the abnormal behavior detection unit 42 acquires detection data for individuals different from each other from the detection data storage unit 34 (step S41). It is determined whether or not there is overlap between both the time period and the range (step S42). As a result of the determination, if there is an overlap (step S42: YES), the abnormal behavior detection unit 42 counts the number of times the two extracted individuals are within a predetermined distance on the same date (step S43).

If the number of times counted in step S43 is equal to or greater than the predetermined value (step S44: YES), it is highly likely that both individuals have been acting together for a certain period of time. It regards and outputs detection information (step S45). As a result, it is considered that at least one individual is in heat and is rubbing against the individuals that often come close to each other in the aisle area, and therefore, it is detected as an estrous behavior. Then, the detection information is transmitted to the user terminal 11 via the communication means 2 by the detection information transmission unit 43 . Thereby, the abnormal behavior of the dairy cow (animal) is sequentially notified to the user.

On the other hand, if the result of determination in step S42 is that there is no overlap (step S42: NO), or if the number of times counted in step S43 is less than a predetermined value (step S44: NO), the abnormal behavior detection unit 42 detects abnormal behavior. Without detection, the process proceeds to step S46. Then, after repeating the above-described processing from step S41 to step S45 for combinations of all individuals in the detection data (step S46: YES), this processing ends.

According to the third embodiment as described above, in addition to the effects of the first embodiment, estrous behavior can be detected based on the time zone and range in which the animal stays in the passage area.

The abnormal behavior detection device 1, the abnormal behavior detection program 1a, and the abnormal behavior detection method according to the present invention are not limited to the above-described embodiments as they are. can be modified and embodied. Also, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the respective embodiments described above. For example, some components may be deleted from all the components shown in each embodiment. Furthermore, components across different embodiments may be combined as appropriate.

For example, in each of the above-described embodiments, as abnormal behavior, estrous behavior and eating loss of dairy cows are detected. Action is fine.

Also, in the above-described second embodiment, an example using a convolutional neural network or a convolutional autoencoder as a deep learning network model has been shown, but the present invention is not limited to these. For example, Variational Auto-Encoder (VAE) may be used, and Structural Similarity (SSIM) may be used as an index for measuring image similarity. Further, as the processing method of the abnormal behavior detection device 1, other techniques may be used as long as they can determine whether the detection data is normal or abnormal.

1 Abnormal Action Detection Device 1a Abnormal Action Detection Program 2 Communication Means 3 Storage Means 4 Arithmetic Processing Means 11 User Terminal 31 Program Storage Section 32 Living Area Data Storage Section 33 Movement Trajectory Data Storage Section 34 Detection Data Storage Section 35 User Data Storage Section 41 detection data creation unit 42 abnormal behavior detection unit 43 detection information transmission unit

Claims (7)

Data indicating the movement trajectory of each individual animal, the movement trajectory data including the date and time and the position information of the animal at the relevant date and time, and the living related to the life of the animal, such as feeding grounds, passages, and beds. a detection data creation unit for creating detection data for detecting abnormal behavior of the animal based on living area data defining location information of the area;
an abnormal behavior detection unit that detects abnormal behavior of the animal based on the detection data;
Abnormal behavior detection device. The detection data creation unit creates detection data including the living area in which the animal stayed, the start date and time of stay in the living area, the end date and time of staying in the living area, and the distance traveled in the living area,
The abnormal behavior detection unit detects that an individual that has moved from an initial feeding ground in the living area to another feeding ground via an aisle area moves a distance that exceeds a predetermined value within an elapsed time that is less than or equal to a predetermined value. 2. The abnormal behavior detection device according to claim 1, which detects as an abnormal behavior when it does. The detection data creation unit creates detection data including a trajectory diagram of the movement of the animal within a predetermined time,
The abnormal behavior detection unit uses a trajectory diagram when abnormal behavior does not occur in the convolutional neural network and a trajectory diagram when abnormal behavior occurs as learning data for a trained neural network that has undergone supervised learning, 2. The abnormal behavior detection device according to claim 1, wherein abnormal behavior is detected by inputting said trajectory diagram. The detection data creation unit creates detection data including a trajectory diagram of the movement of the animal within a predetermined time,
The abnormal behavior detection unit inputs the trajectory diagram to a learned autoencoder which has been subjected to unsupervised learning as learning data when the convolutional autoencoder does not exhibit an abnormal behavior. The abnormal behavior detection device according to claim 1, which detects the The detection data creation unit creates detection data including a time zone in which the animal stayed in the passage area and a range in which the animal stayed in the passage area,
In the abnormal behavior detection unit, in the detection data of different individuals, there is overlap in both the time zone and range of stay in the passage area, and the distance within the predetermined value on the same date and time is more than a predetermined number of times. 2. The abnormal behavior detection device according to claim 1, which detects as an estrus behavior in a case. Data indicating the movement trajectory of each individual animal, the movement trajectory data including the date and time and the position information of the animal at the relevant date and time, and the living related to the life of the animal, such as feeding grounds, passages, and beds. a detection data creation unit for creating detection data for detecting abnormal behavior of the animal based on living area data defining location information of the area;
an abnormal behavior detection unit that detects abnormal behavior of the animal based on the detection data;
anomalous behavior detection program that causes the computer to function. Data indicating the movement trajectory of each individual animal, the movement trajectory data including the date and time and the position information of the animal at the relevant date and time, and the living related to the life of the animal, such as feeding grounds, passages, and beds. a detection data creation step of creating detection data for detecting abnormal behavior of the animal based on living area data defining location information of the area;
an abnormal behavior detection step of detecting abnormal behavior of the animal based on the detection data;
Abnormal behavior detection method.

Images