Body condition diagnosis system for dog, body condition diagnosis method and program

Abstract

To provide a body condition diagnosis system which allows an animal guardian of a dog to easily diagnose a body condition of the dog during walking of the dog.SOLUTION: An animal guardian carries an animal guardian terminal 20 which includes a video function and to which a non-contact thermometer 4 is connected during walking of a dog. A reception part 51 of a server 50 receives from the animal guardian terminal 20, a moving image of the back of the dog which is imaged from an upper side of the dog by the animal guardian during walking of the dog, and data of a road surface temperature measured by the non-contact thermometer 4. A learned model storage part 52 stores a learned model which is generated by machine learning by associating, a body condition index in the moving image with a body condition level of the dog being teacher data. A diagnosis part 53 diagnoses the body condition of the dog on the basis of the learned model, according to the moving image and the road surface temperature received on the reception part 51. A transmission part 54 sends the body condition diagnosis result by the diagnosis part 53 to the animal guardian terminal 20.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a dog health diagnosis system, a dog health diagnosis method, and a dog health diagnosis program. [Background technology]

[0002] Conventionally, systems for managing and diagnosing the health and physical condition of pets such as dogs are known. For example, the system of Patent Document 1 determines the heatstroke risk level of a pet based on a discomfort index calculated from temperature and humidity. The system of Patent Document 2 takes an image of the face of the pet and diagnoses the health condition of the pet from the image data. The system of Patent Document 3 takes a video of the pet and estimates illness based on whether the pet is performing a specific action or not. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2006-345826 A [Patent Document 2] JP 2009-165416 A [Patent Document 3] JP 2014-223063 A Summary of the Invention [Problem to be solved by the invention]

[0004] In Patent Documents 1-3, the target animal is not limited to a dog but is expanded to a "pet", but the present invention focuses on the physical condition diagnosis from the owner's perspective while walking a dog, and the target animal is actually considered to be a dog. Many dog ​​owners have a daily routine of walking their dog, and walking time accounts for a large proportion of communication with the dog. The physical condition of the dog is estimated based on the speed at which the dog walks during the walk and whether the behavior of sniffing roads, telephone poles, etc. is the same as usual. Here, from the viewpoint of the owner holding the leash, only the top of the dog's head and the back including the tail are always visible during the walk, and not much information can be obtained from the face.

[0005] When taking a walk on a hot summer day, it is important to be careful of heatstroke caused by radiant heat from the ground. Although the general level of caution for all dogs can be used as a guideline, there are individual differences, so it is important to take into account the individual differences in the health of each dog. Also, on a daily basis, it is not necessary to find a full-blown illness that requires a veterinarian's diagnosis, but rather to simply judge that the dog is "as energetic as usual" or "a little less energetic than usual."

[0006] The system of Patent Document 1 outputs a general heatstroke risk level for all dogs, and is not adapted to diagnose the physical condition of an individual dog. It is difficult to take a picture of a dog's face and perform a health check using the system of Patent Document 2 while the dog is out for a walk. The system of Patent Document 3 estimates symptoms of serious illnesses such as rhinitis from specific actions in a video, and is not intended for simple physical condition diagnosis during daily walks.

[0007] The present invention has been created in consideration of these points, and its purpose is to provide a health condition diagnosis system, a health condition diagnosis method, and a program that allow a dog owner to easily diagnose the health condition of their dog while walking the dog. [Means for solving the problem]

[0008] The dog physical condition diagnosis system according to the present invention includes, as a server-side configuration, a receiving unit 51, a trained model storage unit 52, a diagnosis unit 53, and a transmitting unit 54. On the client side, the owner carries a pet owner terminal 20 having a video function and connected to a non-contact thermometer 4 while walking the dog.

[0009] The receiving unit receives from the owner terminal a video of the dog's back taken from above by the owner while walking the dog, and road surface temperature data measured by a non-contact thermometer. The trained model storage unit stores a trained model created by machine learning by associating the physical condition index and road surface temperature in the video with the physical condition level of the dog, which is the training data.

[0010] The physical condition indicators in the video include the dog's walking speed, ear movement, tail movement, etc. For example, the learned model stores that the physical condition level tends to decrease when the dog's walking speed is slower than normal, or when the ear or tail movement is slower than normal, or that the physical condition level tends to decrease when the road surface temperature is equal to or higher than a predetermined value.

[0011] The diagnosis unit diagnoses the dog's physical condition based on the trained model in accordance with the video and road surface temperature received by the receiving unit. In other words, the diagnosis unit performs a physical condition diagnosis using AI. The transmission unit transmits the physical condition diagnosis results by the diagnosis unit to the owner's terminal.

[0012] In this dog health diagnosis system, the server uses AI to diagnose the health of the dog based on the video of the dog's back taken by the owner while walking the dog and the road temperature, and sends the diagnosis results to the owner's terminal, making it easy to diagnose the health of each individual dog. Especially during summer walks, the dog's health can be predicted from the road temperature, preventing heatstroke.

[0013] The method for diagnosing a dog's physical condition according to the present invention comprises the following steps. A step (S12) in which the owner carries the owner terminal (20) including a video function and connected to the non-contact thermometer (4) while walking the dog, takes a video of the back of the dog from above, and measures the road surface temperature with the non-contact thermometer. The server receives the video and road surface temperature data transmitted from the owner terminal (S13, S14). A step of diagnosing the dog's physical condition according to the received video and road surface temperature based on a trained model created by machine learning by associating the physical condition indexes and road surface temperature in the video with the physical condition level of the dog, which is the training data (S15). A step of transmitting the physical condition diagnosis result to the owner's terminal (S16)

[0014] The program according to the present invention causes a processor of a server (50) having a function of acquiring a video of a dog's back taken from above by a dog owner carrying a dog owner terminal (20) having a video function and a connected non-contact thermometer (4) while walking the dog, as well as the road surface temperature measured by the non-contact thermometer, and associating the physical condition index and road surface temperature in the video with the dog's physical condition level, which is training data, to execute the following processes. - Processing to receive video and road surface temperature data from the owner's device - Diagnoses the dog's physical condition based on the trained model according to the received video and road surface temperature · Processing to send health check results to the owner's terminal

[0015] The invention relating to the dog health diagnosis method and program also provides the same effects as the invention relating to the dog health diagnosis system described above. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 is a configuration diagram of a dog physical condition diagnosis system according to an embodiment. [Diagram 2] FIG. 13 is a diagram showing an example of a video of a pet's back captured by the owner terminal. [Diagram 3] 13 is a flowchart of a registration and initial learning process. [Figure 4] FIG. 13 is a diagram showing an example of a registration screen. [Diagram 5] FIG. 13 is a diagram showing an example of a data transmission screen. [Figure 6] FIG. 13 is a diagram showing an example of a physical condition level (teacher data) determination screen. [Figure 7] 13 is a flowchart of a physical condition diagnosis process. [Figure 8] FIG. 13 is a diagram showing an example of a diagnosis result screen displayed on the owner terminal. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] (One embodiment) An embodiment of a dog health diagnosis system, a health diagnosis method, and a program according to the present invention will be described with reference to the drawings. Fig. 1 shows the configuration of a health diagnosis system according to an embodiment. In this system, a dog owner terminal 20 (client) carried by a dog owner is connected to a server 50 operated by a system provider via a network NW. The server 50 may be configured as a cloud server.

[0018] In reality, multiple owner terminals 20 are connected to one server 50, but communication between different clients is not assumed. The server 50 transmits and receives information individually to each owner terminal 20 identified by an identification number. Fig. 1 shows only the relationship between one owner terminal 20 and the server 50.

[0019] The owner terminal 20 is mainly a smartphone. Generally, a smartphone includes a camera function for taking still images, as well as a video function for taking videos. In FIG. 1, the owner terminal 20 is shown as having a video 3 attached to it to clearly show that it includes a video function. Note that the reference number "3" is not used when referring to the "video function." A non-contact thermometer 4 is also connected to the owner terminal 20.

[0020] While walking their dog, the owner carries the owner terminal 20 including a video function and connected to the non-contact thermometer 4. Hereinafter, when distinguishing from a general "dog", the "dog" being walked will be referred to as the "beloved dog". While walking their dog, the owner takes a video of the dog's back from above using the video 3 of the owner terminal 20, and measures the temperature of the road surface at the dog's feet using the non-contact thermometer 4. As shown in FIG. 2, the video of the dog's back shows the ears on the top of the head and the tail at the rear end of the body, but not the face. The video is taken for a predetermined period of time (e.g., 5 seconds or more).

[0021] The server 50 has a general hardware configuration including a processor, a ROM, a RAM, a storage, an I / O interface, a display, a communication module, etc. The processor is, for example, a CPU, and controls the operation of the server 50. The storage is, for example, a hard disk drive, and stores programs and data.

[0022] Functionally, the server 50 includes a receiving unit 51, a trained model storage unit 52, a diagnosis unit 53, and a transmission unit 54. The receiving unit 51 and the transmission unit 54 are I / O interfaces. The receiving unit 51 receives from the owner terminal 20 a video of the dog's back taken from above by the owner while walking the dog, and data on the road surface temperature measured by the non-contact thermometer 4. The transmission unit 54 transmits the results of the physical condition diagnosis by the diagnosis unit 53 to the owner terminal 20.

[0023] The trained model storage unit 52 and the diagnosis unit 53 constitute an AI that performs machine learning through deep learning. The trained model storage unit 52 stores a trained model created by machine learning using training data. This model is created by associating the physical condition index and road surface temperature in the video with the physical condition level of the dog, which is the training data. The "physical condition index in the video" is information that serves as an index for estimating the physical condition of the dog during a walk, and will be described in detail later.

[0024] The diagnosis unit 53 diagnoses the dog's physical condition based on the trained model in accordance with the video and the road surface temperature received by the receiving unit 51. In other words, the diagnosis unit performs a physical condition diagnosis using AI.

[0025] (Registration / Initial Learning) The registration and initial learning process will be described with reference to the flowchart in Fig. 3 and Figs. 4 to 6 showing examples of the owner terminal screen. The symbol "S" in the flowchart means a step. S01 corresponds to the registration process, and S02 to S06 correspond to the initial learning process. The initial learning process is a process in which the owner inputs the physical condition level of their dog, which they keep track of on a daily basis, as training data for the AI.

[0026] In S01, the owner registers information about his / her beloved dog. As shown in FIG. 4, the name, date of birth, sex, and breed of the dog are input in the information input section 211 of the registration screen 21. The gender can be selected and input as male or female, and the breed can be selected and input as the main breed (for example, Yorkshire terrier). The correction button 215 is used when correcting the input information, and the confirmation button 216 is used when confirming the input information. After confirming the input information, the owner presses the registration button 217, and the dog information of each owner is registered in the server 50.

[0027] In S02, while walking the dog, the owner takes a video of the dog's back from above using the video 3 of the owner terminal 20, and measures the road surface temperature at the dog's feet using the non-contact thermometer 4. As shown in the example of Fig. 5, the data transmission screen 22 displays an image from the video 3 in the video display section 223 together with the current time, and the road surface temperature (e.g., 33.0°C) in the temperature display section 224.

[0028] The Modify button 225 is used to modify data, and the Confirm button 226 is used to confirm data. During initial learning, after confirming the data, when the owner presses the physical condition level learning button 227, the screen transitions to the physical condition level judgment screen 23 in Fig. 6. The Diagnosis button 228 is used in the physical condition diagnosis process described later.

[0029] In S03, the owner subjectively judges the physical condition level of the dog at that time, that is, in the situation when the video and road surface temperature data were obtained. As shown in Fig. 6, the physical condition level judgment screen 23 displays radio buttons 231 to 234 for judging the physical condition level in four stages: good, questionable, slightly poor, and poor. Any one of the radio buttons 231 to 234 can be selected. Note that in other embodiments, the number of stages of the physical condition level is not limited to four.

[0030] "Good" means the dog is in its usual healthy state. "Poor" means the dog is clearly not in good health and it would be a good idea to take it to the vet. "Slightly unhealthy" means the dog seems a little lethargic, but it is best to wait and see whether to take it to the vet. "A little poor" means the dog is in good health, but may be feeling stressed due to the presence of other dogs around. Owners can decide the criteria for each health level.

[0031] After determining the physical condition level using radio buttons 231 to 234, the owner presses send button 239. As a result, in S04, the owner transmits a set of the video and road surface temperature data and the physical condition level determined at that time to server 50. The physical condition level determined by the owner is treated as training data for an AI learning model.

[0032] The initial learning from S02 to S04 is repeated for several days. If it is determined in S05 that a predetermined amount of training data has been accumulated (YES), the process proceeds to S06.

[0033] The AI ​​evaluates the physical condition indicators appearing in the video of the dog's back by image analysis. The dog's walking speed, ear movement, tail movement, etc. are analyzed as physical condition indicators. For example, the walking speed and agility of ear and tail movement are recognized as patterns. In S06, the AI ​​performs machine learning by associating the physical condition indicators and road surface temperature in the video with the dog's physical condition level, which is the teacher data, to create a trained model. The created trained model is stored in the trained model storage unit 52.

[0034] However, in reality, the physical condition level is "good" almost every day, and at most occasionally becomes "slight," so it is unlikely that teacher data for "slightly delinquent" or "delinquent" can be obtained. Therefore, teacher data for "slightly delinquent" or "delinquent" may be mechanically created by image processing of the video, for example, intentionally slowing down the walking speed or intentionally slowing down the movement of the ears and tail.

[0035] (Health check) The physical condition diagnosis process performed after the trained model is created will be described with reference to the flowchart in Fig. 7 and Fig. 8 showing an example of a screen of the owner terminal 20. For the screen of the owner terminal 20, Fig. 5 and Fig. 6 referred to in the initial learning process will also be referred to again.

[0036] S12 is the same as S02 in the initial learning process. While walking the dog, the owner takes a video of the dog's back from above using the video 3 of the owner terminal 20, and measures the road surface temperature at the dog's feet using the non-contact thermometer 4. The video and road surface temperature are displayed on the data transmission screen 22 in Fig. 5, and the process is the same as the initial learning process until the data is confirmed using the Confirm button 226.

[0037] When the owner presses the diagnosis button 228 on the data transmission screen 22 in S13, data is transmitted from the owner terminal 20 to the server 50. In S14, the receiving unit 51 receives the video and road surface temperature data transmitted from the owner terminal 20. The diagnosing unit 53 calls up the trained model stored in the trained model memory unit 52 for each registered dog, based on the identification number of the owner terminal 20.

[0038] In S15, the diagnosis unit 53 diagnoses the dog's physical condition based on the learned model and in accordance with the received video and road surface temperature. Specifically, the diagnosis unit 53 recognizes the movement (displacement) of each part of the dog's body in the video as a physical condition index using a convolutional neural network (CNN). The walking speed is calculated from the relative speed of the torso to a reference position in the background. In addition, patterns such as the displacement range, trajectory, displacement speed, etc. of the ears based on the top of the head, and the displacement range, trajectory, displacement speed, etc. of the tail based on the torso are recognized.

[0039] Also, when taking a walk on a hot summer day, it is important to be careful of heatstroke caused by rising road surface temperature. Even if the air temperature is the same, the road surface temperature tends to be higher in the afternoon when heat accumulates on the road than in the morning. Therefore, in the health diagnosis system of this embodiment, the road surface temperature measured by the non-contact thermometer 4 is used as a parameter for health diagnosis, rather than the air temperature. The diagnosis unit 53 comprehensively judges the health index and the road surface temperature in the video to diagnose the dog's health. Since symptoms of heatstroke may appear later, even if the health index at that time is good, it is desirable to diagnose the health level to the down side, including the meaning of prevention, when the road surface temperature is high.

[0040] In S16, the transmission unit 54 transmits the result of the physical condition diagnosis by the diagnosis unit 53 to the owner terminal 20. As exemplified in Fig. 8, the diagnosis result screen 24 displays the physical condition level and message of the diagnosis result together with the current time on the result display unit 241. If the physical condition level is "good", "Your dog's physical condition is good. Please continue to enjoy your walks" is displayed.

[0041] Other messages that will be displayed include "Is there anything around your pet that is causing stress?" if your pet's health level is "bad?", "Try to limit walks and keep an eye on your pet's condition" if your pet's health level is "poor", and "Stop walks and take your pet to the vet" if your pet's health level is "poor".

[0042] If the owner who has viewed the diagnosis result screen 24 does not agree with the physical condition level diagnosed by the AI, the owner can also modify the physical condition level. In S17, it is determined whether the physical condition level needs to be modified, and if the answer is YES, in S18, the owner transmits the modified physical condition level from the owner terminal 20 to the server 50. At this time, if the owner presses the physical condition level modification button 247 on the diagnosis result screen 24, the screen transitions to the physical condition level judgment screen 23 (FIG. 6), where the owner judges the modified physical condition level and transmits it to the server 50. If the physical condition level needs not to be modified (S17: NO), the owner presses the diagnosis completion button 248 on the diagnosis result screen 24.

[0043] The corrected physical condition level becomes new teacher data. Even if there is no correction, the accuracy of machine learning is improved by adding the new diagnosis result to the trained model. Thus, in S19, the trained model storage unit 52 updates and stores the trained model.

[0044] The flowchart in FIG. 7 shows a method for diagnosing a dog's physical condition, and also shows a program for causing the processor of the server 50 to execute the following processes. Processing for receiving video and road surface temperature data from the owner terminal 20 - Diagnoses the dog's physical condition based on the trained model according to the received video and road surface temperature Processing for transmitting physical condition diagnosis results to the owner terminal 20

[0045] In the dog health diagnosis system and method of this embodiment, while the owner is walking the dog, the owner can take a video of the dog's back from the owner's viewpoint and measure the road surface temperature to diagnose the dog's health. Compared to the system of Patent Document 2, where it is difficult to take an image of the dog's face, it is easy to take a video of the dog's back.

[0046] Although this embodiment shares with the system of Patent Document 3 the use of video for diagnosis, it differs from the system of Patent Document 3 in that it also diagnoses the dog's physical condition from the perspective of heat stroke based on the road surface temperature. Also, while the system of Patent Document 1 determines a discomfort index based on air temperature and humidity, this embodiment uses the road surface temperature, making it possible to more accurately reflect the effect on the dog's physical condition.

[0047] In addition, while the system of Patent Document 3 and the like is intended to detect serious illnesses such as rhinitis, the system of this embodiment is intended to make simple judgments such as more everyday conditions such as "as healthy as usual" or "slightly less healthy than usual." For at least "good" and "slightly poor," the health level judged by the owner himself, who is in contact with his dog every day, becomes the teacher data. Therefore, it is possible to manage the health of each dog while taking into account the individual differences.

[0048] If a trained model is created based on training data of the owner's own judgment of the dog's physical condition, there seems to be little benefit in having an AI diagnose the dog's physical condition. However, the owner's perception may change from a few months ago. Therefore, by storing the owner's past judgment results as training data in the AI, the judgment criteria for the dog's physical condition level can be maintained within a certain range regardless of changes in the owner's perception.

[0049] As described above, the present invention is not limited to the above-described embodiment, and can be embodied in various forms without departing from the spirit of the present invention. [Explanation of symbols]

[0050] 20 Owner terminal, 3 Video, 4 Non-contact thermometer, 50 servers, 51: Receiving unit; 52: Trained model storage unit; 53 ··· diagnostic unit, 54 ··· transmission unit.

Claims

1. a receiving unit (51) for receiving from the owner terminal (20) including a video function and connected to a non-contact thermometer (4) a video of the dog's back taken from above the dog by the owner carrying the dog while walking the dog, and data on the road surface temperature measured by the non-contact thermometer; a trained model storage unit (52) that stores a trained model created by machine learning by associating the physical condition index and the road surface temperature in the video with the physical condition level of the dog, which is training data; a diagnosis unit (53) that diagnoses the physical condition of the dog based on the trained model in response to the video and the road surface temperature received by the receiving unit; A transmission unit (54) that transmits the physical condition diagnosis result by the diagnosis unit to the owner terminal; A dog health diagnosis system equipped with

2. A step (S12) in which a dog owner carries a dog owner terminal (20) including a video function and connected to a non-contact thermometer (4) while walking the dog, takes a video of the dog's back from above, and measures the road surface temperature with the non-contact thermometer; A step (S13, S14) in which a server receives the video and the road surface temperature data transmitted from the owner terminal; A step (S15) of diagnosing the dog's physical condition according to the received video and the road surface temperature based on a trained model created by machine learning by associating the physical condition index and the road surface temperature in the video with the physical condition level of the dog, which is training data; A step (S16) of transmitting the physical condition diagnosis result to the owner terminal; A method for diagnosing a dog's health condition, including:

3. A processor of a server (50) having a function of acquiring a video of the back of a dog taken from above the dog by a dog owner carrying a dog owner terminal (20) including a video function and connected to a non-contact thermometer (4) while walking the dog, and a road surface temperature measured by the non-contact thermometer, and storing a trained model created by machine learning by associating the physical condition index and the road surface temperature in the video with the physical condition level of the dog, which is training data; receiving the video and the road surface temperature data from the owner terminal; A process of diagnosing the dog's physical condition based on the trained model in response to the received video and the road surface temperature; A process of transmitting the physical condition diagnosis result to the owner terminal; A program that executes the following.

Images