WO2024040367A1

WO2024040367A1 - Method and apparatus for monitoring organisms by means of artificial intelligence-based automatic system

Info

Publication number: WO2024040367A1
Application number: PCT/CN2022/113817
Authority: WO
Inventors: 王士修; 武玟嫒
Original assignee: 探极生物科技有限公司
Priority date: 2022-08-22
Filing date: 2022-08-22
Publication date: 2024-02-29

Abstract

A method and apparatus for monitoring organisms by means of an artificial intelligence-based automatic system. According to the method and apparatus, an autonomous moving apparatus with artificial intelligence and edge computing functions is arranged beside a frame for an organism rearing cage, facilitating the detection and management of the condition of experimental animals in the organism rearing cage; data operation or analysis is performed on captured organism images, and meanwhile, the organism images are transmitted back to a central management platform, so that the purposes of automatic monitoring and care management of organisms are achieved, and then the conditions in the organism rearing cage are standardized, thereby reducing the interference with the lives of the experimental animals so as to improve the care quality of the experimental animals. Therefore, in addition to solving the problem of manual ward rounds, the method and apparatus can improve the anomaly detection rate and pollution control level, thereby achieving the purposes of automatic monitoring and care management of organisms.

Description

Methods and devices for monitoring organisms by automated systems based on artificial intelligence

Technical field

The present invention relates to a method and device for monitoring organisms, and in particular to a method and device for monitoring organisms by an automated system based on artificial intelligence.

Background technique

According to the current situation, experimental animal care facilities are facing three major problems:

1. The artificial care mode of experimental animals is time-consuming and the detection status is not good: the daily care of experimental animals is carried out manually to observe the animal status. Experimental animals need to undergo routine care and observation once a day in the morning and afternoon. The observation content includes counting animals. Abnormal conditions such as the number of animals, whether the animals are injured or dead, and whether the living environment of the animals is suitable. The current manual method of patrolling the rooms and observing cage by cage is not only time-consuming, but also has a low abnormality detection rate due to the large number of animal cages. It is often impossible to immediately reflect the conditions of the animals in the cages, resulting in animals being in a state of emergency and unable to implement animal welfare.

2. Animal room pollution control and light cycle considerations: Due to pollution control in the animal room, personnel entering the animal room need to dress before entering the animal room. Therefore, staff cannot continuously observe animal activities for a long time, even if they can observe the animals in the animal room. It is also very limited, and there has been a lack of long-term and continuous animal observation data. Another consideration is to simulate the animals' normal daily routine. The lighting in the animal house is automatically controlled for 12 hours of light and 12 hours of darkness. Entering the animal house at night will not affect the animal's light cycle, so special equipment and training are required to enter. As a result, night care cannot be carried out and data on the activity patterns of experimental animals at night cannot be accumulated.

3. The conditions in animal cages cannot be standardized: Because the inspection work is performed manually, everyone has different standards for judging the conditions in the cage, which often leads to no unified standards for the conditions in the same animal cage. Therefore, when changing cages, the same animal room is often used to complete the cage changing work at the same time. However, animals do not like to be disturbed by nature. Every cage changing action is an urgent stimulus for the animals. If the cage changing intervention can be reduced, it can Taking into account the quality of life of experimental animals is the best model for experimental animal care.

Therefore, after active thinking, prototype testing and continuous improvement, the inventor of this case finally developed a simple and practical improvement method, which can especially solve the above-mentioned shortcomings.

Contents of the invention

The purpose of the present invention is to provide an automated system-based biological monitoring method based on artificial intelligence, which can overcome the shortcomings of the existing technology, effectively solve the problem of manual room inspections, improve the abnormal detection rate and pollution control level, and can be automated. Monitor and care for managed organisms.

In order to achieve the above objectives, the present invention discloses a method for monitoring organisms by an automated system based on artificial intelligence, which is characterized by comprising the following steps:

Step 1: Acquire multiple image data through an autonomous mobile device, which includes a host unit and a sensing unit, to form an image data set, and the autonomous mobile device is placed next to a biological breeding cage;

Step 2: For the image data of the image data group, capture the required image frames and convert them into static images through the motion detection algorithm and the key frame capture algorithm;

Step 3: Use a computer equipped with a graphical user interface software to mark the coordinates of the target object in the above static image through the graphical user interface software to form a mark data, and store the mark data in The computer, with the aid of the graphical algorithm of the graphical user interface software, quickly completes labeling of the labeled data to form a labeled data set for training of a machine learning machine;

Step 4: Input the data of the marked data group into the machine learning machine, and establish a recognition model through the machine learning algorithm;

Step 5: Deploy the identification model to the host unit;

Step 6: Process the new image data acquired by the sensing unit of the autonomous mobile device through steps 2 and 3 to form a new label data group, and input it into the machine learning machine. Through the above By comparing and analyzing the machine learning algorithm with the identification model, the comparison and identification results can be obtained to identify the status of the organism and its feeding environment. At the same time, the comparison and identification results are transmitted to a central management platform for monitoring and abnormal notification. Source of information, thereby achieving the purpose of automated monitoring, care and management of organisms.

Wherein, the autonomous mobile device is a linear track group, a self-propelled vehicle or a drone, and the host unit is an edge computing host. The edge computing host includes a central processing unit, a memory, a graphics computing unit, a peripheral I/O interface, Wireless transmission, data storage unit and power supply unit. The sensing unit includes a camera module, infrared detection, temperature meter, humidity meter, directional microphone, vibration meter and pressure meter or any combination of the above.

Among them, the image data set includes animal species, environmental conditions and animal behaviors, and the image data set is further normalized through computer vision algorithms.

Among them, the marking data set includes an object recognition module, an image segmentation module, an animal entity recognition module and an animal behavior recognition module.

Among them, the object recognition module marks the location of the marked data in the form of bounding boxes one by one, and stores the marking data in the object recognition module. The image segmentation module records the objects belonging to the marked data one by one. The pixels in the contour, object body and background areas are marked, and the annotation data is stored in the image segmentation module. The animal entity recognition module identifies the facial features belonging to the object's face image in the marked data one by one. The area of the point is marked, and the marking data is stored in the animal entity recognition module. The animal behavior recognition module marks the target joint points belonging to the object in the marked data on the image one by one, and stores the marking data. in the animal behavior recognition module.

Among them, in step 6, the above-mentioned identification results of some abnormalities are reviewed by the administrator to evaluate whether they need to be re-labeled and classified to form a new label data group so that the new label data group can be continuously input to the machine learning The machine is used to retrain the recognition model to improve the accuracy of the recognition model.

Also disclosed is an automated system based on artificial intelligence to monitor biological devices, which is installed on a biological breeding cage. The biological breeding cage is equipped with a plurality of biological breeding cages. It is characterized in that the artificial intelligence-based automated system monitors biological The device includes:

A track unit, which includes at least one track group, the track group has a first track, the first track is provided on one side of the biological breeding cages;

A driving unit electrically connected to the track unit, the driving unit is provided with at least one driver corresponding to the track group, and the driver drives the track group to operate;

A host unit. The host unit is provided with at least one host corresponding to the first track. The bottom of the host is provided with a first set of connecting plates. The first set of connecting plates is provided with a second set of connecting plates. The second set of connecting plates The board is located in this track group;

A sensing unit is provided with a sensor corresponding to the host. The sensor is electrically connected to the host and is located at one end of the host.

Among them, the host is an edge computing host, including a central processing unit, a memory, a graphics computing unit, a peripheral I/O interface, a wireless transmission and a data storage unit, and the sensor includes a camera module, infrared detection, and a thermometer. , hygrometer, directional microphone, vibration meter and pressure meter or any combination of the above.

Among them, one end of the main machine is connected to an adjusting frame. The adjusting frame includes a front frame body and a rear frame body. Both sides of the front frame body are connected to the rear frame body by a first fixing member, and the rear frame The body is connected to the main machine by a plurality of second fixing members.

Wherein, the first track is arranged horizontally or vertically.

It also includes a support unit, which is provided with at least one support frame corresponding to the first track. One end of the support frame is fixed on the biological breeding cage frame, and the other end is fixed on the first track. In addition, the support frame The support unit also includes a supporting frame, which is erected around the biological breeding cage or on one side of the biological breeding cage. The supporting frame is composed of a plurality of supporting rods, so that the supporting frame The overall frame is in an inverted U shape or square.

Among them, the track set has a first track, two second tracks and a synchronous round rod. The second tracks are located on both sides of the support frame and are respectively fixed to the corresponding support rods. The driver is located on One of the second rails, the first rail is fixed to each of the second rails by two fixing brackets, and the two ends of the synchronized round rod are respectively fixed to each of the second rails.

It also includes a power supply unit, the power supply unit is provided with a power supply module corresponding to the host, and the power supply module provides the power required by the host and the sensor.

Among them, the first set of connecting plates is provided with a plurality of slide blocks, and the second set of connecting plates is provided with a plurality of slide grooves corresponding to the slide blocks of the first set of connecting plates, and the slide blocks are fastened to the These chutes allow the main unit to be placed on the first track and move on the first track.

Among them, the first set of connecting plates is provided with a plurality of magnetic suction parts, and the second set of connecting plates is provided with a plurality of magnetic suction slots corresponding to the magnetic suction parts of the first set of connecting plates. The components are adsorbed on the magnetic slots, so that the main unit is placed on the first track and moves on the first track.

Among them, the first set of connecting plates are respectively provided with a pressing piece on both sides, and the second set of connecting plates are provided with two pressing holes corresponding to the two pressing pieces of the first set of connecting plates, and are clamped on the These pressing holes allow the main unit to be placed on the first track and move on the first track.

Among them, the first set of connecting plates are respectively provided with a buckle on both sides, and the second set of connecting plates are provided with two buckling holes corresponding to the two buckling parts of the first set of connecting plates. The fasteners are fastened to the buckle holes, so that the main unit is placed on the first track and moves on the first track.

With the above method and structure, the present invention allows the main machine to reciprocate laterally or axially on the first track, so that the sensors provided on the main machine can monitor the activity status of the experimental animals in the biological breeding cages for a long time and continuously. Observation, further standardizing the conditions in the cages of these organisms, and only replacing the cages of the organisms that need to be replaced, reducing interference to the lives of experimental animals, and improving the quality of experimental animal care. Therefore, in addition to solving the problem of manual rounds In addition to solving problems, it can also improve the abnormal detection rate and pollution control level, achieving the purpose of automatically monitoring and caring for and managing organisms.

Description of drawings

Figure 1 is a schematic flow diagram of the present invention.

Figure 2 is a schematic diagram of an object recognition module used in the present invention.

FIG. 3 is a schematic diagram of the image segmentation module of the present invention.

Figure 4 is a schematic diagram of the animal entity recognition module of the present invention.

Figure 5 is a schematic diagram of the animal behavior recognition module of the present invention.

Figure 6 is a perspective view of the first embodiment of the present invention.

Figure 7 is a partial enlarged view of the first embodiment of the present invention.

FIG. 8 is a schematic diagram of the main unit assembled on the first track according to the first embodiment of the present invention.

Figure 9 is a partially exploded view of the first embodiment of the present invention.

FIG. 10 is a schematic diagram of the sensor operating on the host according to the first embodiment of the present invention.

FIG. 11 is a schematic diagram of the host machine reciprocating on a horizontally disposed first track according to the first embodiment of the present invention.

FIG. 12 is a schematic diagram of the host machine reciprocating on the first rail arranged longitudinally according to the first embodiment of the present invention.

Figure 13 is a perspective view of the second embodiment of the present invention.

Figure 14 is a schematic diagram of the first track reciprocating up and down according to the second embodiment of the present invention.

Figure 15 is a schematic diagram of the first track reciprocating left and right according to the second embodiment of the present invention.

FIG. 16 is a schematic diagram of the host moving left and right on the first rail and up and down on the second rail according to the second embodiment of the present invention.

Figure 17 is a schematic diagram of the main unit assembled on the first track according to the third embodiment of the present invention.

Figure 18 is a schematic diagram of the main unit assembled on the first track according to the fourth embodiment of the present invention.

FIG. 19 is a schematic diagram of the main unit assembled on the first track according to the fifth embodiment of the present invention.

Detailed ways

Please refer to Figure 1 and Figure 6, which discloses a method of biological monitoring based on an automated system based on artificial intelligence, which includes the following steps:

Step 1S1: Acquire multiple image data through an autonomous mobile device to form an image data set, and the autonomous mobile device is placed next to a biological breeding cage 200.

Among them, the image data can be static images or dynamic videos.

The autonomous mobile device can be a linear track set, a self-propelled vehicle or a drone. The autonomous mobile device includes a host unit 40 and a sensing unit 50. The host unit 40 is an edge computing host. The edge The computing host includes a central processing unit (CPU), memory, graphics processing unit (GPU, Graphics processing unit), peripheral I/O interface, wireless transmission, data storage unit and power supply unit. The sensing unit 50 includes a camera module , infrared detection, temperature meter, humidity meter, directional microphone, vibration meter and pressure meter or any combination of the above.

Among them, the image data set includes animal species (such as age, body color), environmental conditions (such as skewed cage frames, low feed stocks, wet bedding due to excrement or leaking drinking fountains) and animal behaviors (such as eating feed, Climbing, fighting, mating, abnormal mobility).

Step 2S2: For the image data of the image data group, through the motion detection algorithm (motion detection algorithm) and the key frame extraction algorithm (key frames extraction algorithm), capture some frames with reference value and convert them into Static images, and images generated by different models of camera modules can be further normalized through computer vision algorithms to enhance data consistency and help improve a machine learning machine. Training speed and accuracy.

Step 3S3: As shown in Figures 2 to 5, use a computer equipped with a graphical user interface software (GUI, graphic user interface) to mark the target in the above static image through the graphical user interface software The coordinates of the object form a mark data, and the mark data is stored in the computer. In the embodiment of the present invention, the above-mentioned target objects are experimental animals and objects in the environment that need to be marked, and the computer can be a desktop computer. Or a laptop computer, which can also be used with the graphical algorithm of the graphical user interface software to assist the image annotator to complete the annotation of the tagged data in a faster way to form a tagged data group for the machine learning machine. For training purposes, the marking data set includes an object recognition module, an image segmentation module, an animal entity recognition module and an animal behavior recognition module. Users can select based on different above-mentioned target objects or different monitoring purposes. The module to use.

Among them, the object recognition module uses the graphical user interface software to mark the location of the marking data one by one in the form of bounding box A, as shown in Figure 2, and stores the marking data in the In the object recognition module, the image segmentation module uses the graphical user interface software to identify the areas belonging to the object outline (boundary/contour) B, object body (object) C and background (background) D of the marked data one by one. The pixels (image pixels) are marked, as shown in Figure 3, and the labeling data is stored in the image segmentation module. The animal entity recognition module uses the graphical user interface software to classify the labeling data one by one. The area belonging to facial landmark points (facial landmark) P1 to P7 on the object's facial image is marked, as shown in Figure 4, and its annotation data is stored in the animal entity recognition module, and the animal behavior recognition module The group uses the graphical user interface software to mark out the target joint points (joints points or decision points) E belonging to the object in the mark data on the image one by one, as shown in Figure 5, and stores the mark data in the animal In the behavior recognition module.

Step 4S4: Input the data of the marked data group into the machine learning machine, and establish a recognition model through the machine learning algorithm.

Step 5S5: Deploy the recognition model to the host unit 40.

Step 6S6: Process the new image data acquired by the sensing unit 50 of the autonomous mobile device through steps 2 and 3 to form a new label data set, and input it into the machine learning machine. The above-mentioned machine learning algorithm is compared and analyzed with the identification model to obtain a comparison and identification result that identifies the organism and its feeding environment status. At the same time, the above-mentioned comparison and identification result is transmitted to a central management platform, where the central management platform It can be a local computer or a cloud virtual host as a source of information for monitoring and abnormal notification, thereby achieving the purpose of automatically monitoring and caring for and managing living things.

Among them, the machine learning algorithm in step 4 can vary according to application requirements, ranging from rule-based calculations, such as clustering or support vector machine (SVM), etc. to learning-based algorithms, and the latter is mainly based on deep learning (Deep Learning) with neural networks as the core architecture.

Referring to FIGS. 6 to 10 , the present invention further discloses a first embodiment of a biological monitoring device 100 based on an automated system based on artificial intelligence, which is installed on a biological breeding cage 200 . The biological breeding cage 200 Equipped with a plurality of biological breeding cages 210, the artificial intelligence-based automated system monitoring biological device 100 includes:

A track unit 10 includes at least one track group 11. The track group 11 has a first track 12. The first track 12 is provided on one side of the biological breeding cages 210. The first track 12 can be horizontal. setting or portrait setting.

A driving unit 20 is electrically connected to the track unit 10. The driving unit 20 is provided with a driver 21 corresponding to the first track 12. In the embodiment of the present invention, the driver 21 can be a motor or a pneumatic cylinder. The driver 21 is provided at one end of the first rail 12, and the driver 21 can drive the first rail 12 to operate.

A support unit 30. The support unit 30 is provided with at least one support frame 31 corresponding to the first track 12. One end of the support frame 31 is fixed to the biological breeding cage 200, and the other end is fixed to the first Track 12.

A host unit 40. The host unit 40 is provided with at least one host 41 corresponding to the first track 12. The host 41 is located on the first track 12. The host 41 is an edge computing host. The edge computing host includes a central Processor (CPU), memory, graphics processing unit (GPU, Graphics processing unit), peripheral I/O interface, wireless transmission and data storage unit, and the bottom of the host 41 is provided with a first set of connecting boards 42, the first The set of connecting plates 42 is provided with a second set of connecting plates 43. The second set of connecting plates 43 is located on the first track 12. In the embodiment of the present invention, the first set of connecting plates 42 is provided with a plurality of sliders 421. The second set of connecting plates 43 is provided with a plurality of slide grooves 431 corresponding to the slide blocks 421 of the first set of connecting plates 42. The slide blocks 421 are fastened to the slide grooves 431, so that the main unit 41 is assembled on the first rail 12 and can move on the first rail 12. Another end of the main machine 41 is connected to an adjusting frame 44. The adjusting frame 44 includes a front frame 45 and a rear frame 46. , both sides of the front frame 45 are connected to the rear frame 46 by a first fixing part 47, and the rear frame 46 is connected to the main unit 41 by a plurality of second fixing parts 48, wherein the transparent By loosening the first fixing parts 47 , the front frame body 45 can be rotated relative to the rear frame body 46 , or by loosening the second fixing parts 48 , the rear frame body 46 can be rotated relative to the main unit 41 Adjust the position up or down.

A sensing unit 50. The sensing unit 50 is provided with a sensor 51 corresponding to the host 41. The sensor 51 includes a camera module, infrared detection, a thermometer, a humidity meter, a directional microphone, and a vibration meter. and a pressure gauge or any combination of the above. The sensor 51 is electrically connected to the host 41 and is located on the front frame 45 of the host 41 .

A power supply unit 60 is provided with a power supply module 61 corresponding to the host 41 . The power supply module 61 can provide the power required by the host 41 and the sensor 51 .

Referring to Figures 11 and 12 in conjunction with Figure 7, the driver 21 drives the first rail 12 to rotate and drives the main machine 41 to reciprocate laterally or axially on the first rail 12 through the main machine 41. The sensor 51 provided on the host 41 can observe the activity status of the experimental animals in the biological breeding cages 210 for a long time and continuously, and further standardize the conditions in the biological breeding cages 210, only for those that need to be replaced. The biological breeding cage 210 is replaced to reduce interference to the life of experimental animals and improve the quality of experimental animal care. This not only solves the problem of manual rounds, but also improves the abnormality detection rate and pollution control level.

Referring to FIGS. 13 to 16 , a second embodiment of the device of the present invention is disclosed. The difference between the second embodiment of the present invention and the aforementioned first embodiment is that the support unit 30 also includes a supporting frame 32 , the supporting frame 32 can be erected around the biological breeding cage 200 or on one side of the biological breeding cage 200. The supporting frame 32 is composed of a plurality of supporting rods 33, so that the supporting frame 32 can 32 can be slightly inverted U-shaped or square as a whole. The track group 11 has one first track 12, two second tracks 13 and a synchronous round rod 14. The second tracks 13 are located on the support frame 32. Both sides are fixed to the corresponding support rods 33 respectively. The driver 21 is arranged on one of the second rails 13. The first rail 12 is fixed to each of the second rails 13 by two fixing brackets 34, and the synchronization Both ends of the round rod 14 are respectively fixed to the second rails 13. Therefore, the driver 21 drives the corresponding second rail 13 and the synchronous round rod 14 to rotate at the same time, so that the synchronous round rod 14 drives the other synchronous round rod 14. The second rail 13 is activated, thereby causing the fixing brackets 34 to move synchronously and in the same direction to link the first rail 12 , so that the first rail 12 and the main machine 41 move along the long axis direction of the second rails 13 For reciprocating movement, another driver 21 may be further provided on the first rail 12 so that the other driver 21 drives the first rail 12 to rotate at the same time, so that the host 41 can move on the first rail 12 at the same time to lift. Convenience in use of the present invention.

Referring to Figure 17, a third embodiment of the device of the present invention is disclosed. The difference between the third embodiment of the present invention and the aforementioned first embodiment is that the first assembly plate 42 is provided with a plurality of magnetic attractions. 422, and the second set of connecting plates 43 is provided with a plurality of magnetic slots 432 corresponding to the magnetic pieces 422 of the first set of connecting plates 42, and is attracted to the magnetic pieces through the magnetic pieces 422. The slot 432 allows the host 41 to be assembled on the first rail 12 and move on the first rail 12 .

Referring to Figure 18, a fourth embodiment of the device of the present invention is disclosed. The difference between the fourth embodiment of the present invention and the aforementioned first embodiment is that the first assembly plate 42 is provided with a push button on both sides. 423, and the second set of connecting plates 43 is provided with two pressing holes 433 corresponding to the two pressing parts 423 of the first set of connecting plates 42, and the pressing parts 423 are clamped in the pressing holes 433, so that The host 41 is assembled on the first rail 12 and can move on the first rail 12 .

Referring to Figure 19, a fifth embodiment of the device of the present invention is disclosed. The fifth embodiment of the present invention is different from the first embodiment in that a card is provided on both sides of the first assembly plate 42. Fasteners 424, and the second set of connecting plates 43 are provided with two buckling holes 434 corresponding to the two fastening parts 424 of the first set of connecting plates 42, and are fastened to the buckles through the fastening parts 424. The button holes 434 enable the main unit 41 to be assembled on the first rail 12 and move on the first rail 12 .

Claims

A method for monitoring organisms with an automated system based on artificial intelligence, characterized by comprising the following steps:

Step 1: Acquire multiple image data through an autonomous mobile device, which includes a host unit and a sensing unit, to form an image data set, and the autonomous mobile device is placed next to a biological breeding cage;

Step 2: For the image data of the image data group, capture the required image frames and convert them into static images through the motion detection algorithm and the key frame capture algorithm;

Step 3: Use a computer equipped with a graphical user interface software to mark the coordinates of the target object in the above static image through the graphical user interface software to form a mark data, and store the mark data in The computer, with the aid of the graphical algorithm of the graphical user interface software, quickly completes labeling of the labeled data to form a labeled data set for training of a machine learning machine;

Step 4: Input the data of the marked data group into the machine learning machine, and establish a recognition model through the machine learning algorithm;

Step 5: Deploy the identification model to the host unit;

Step 6: Process the new image data acquired by the sensing unit of the autonomous mobile device through steps 2 and 3 to form a new label data group, and input it into the machine learning machine. Through the above By comparing and analyzing the machine learning algorithm with the identification model, the comparison and identification results can be obtained to identify the status of the organism and its feeding environment. At the same time, the comparison and identification results are transmitted to a central management platform for monitoring and abnormal notification. Source of information, thereby achieving the purpose of automated monitoring, care and management of organisms.
The method of monitoring living things by an automated system based on artificial intelligence as claimed in claim 1, wherein the autonomous mobile device is a linear track set, a self-propelled vehicle or a drone, the host unit is an edge computing host, and the The edge computing host includes a central processing unit, memory, graphics computing unit, peripheral I/O interface, wireless transmission, data storage unit and power supply unit. The sensing unit includes a camera module, infrared detection, temperature meter, and humidity meter. , directional microphones, vibrators and pressure gauges or any combination of the above.
The method of biological monitoring based on artificial intelligence automated system as claimed in claim 1, characterized in that the image data set includes animal species, environmental conditions and animal behavior, and the image data set is further processed through computer vision algorithms. Image normalization.
The method of biological monitoring by an automated system based on artificial intelligence as claimed in claim 1, wherein the marking data set includes an object recognition module, an image segmentation module, an animal entity recognition module and an animal behavior Identify the module.
The method of biological monitoring by an automated system based on artificial intelligence according to claim 4, characterized in that the object recognition module uses the graphical user interface software system to mark the location of the marked data one by one in a bounding box manner. The location and the labeling data are stored in the object recognition module. The image segmentation module uses the graphical user interface software to mark out the pixels of the object outline, object body and background area belonging to the labeling data one by one. , and the labeling data is stored in the image segmentation module. The animal entity recognition module uses the graphical user interface software to label the areas belonging to the facial feature points on the object's face image in the labeling data one by one. , and the labeling data is stored in the animal entity recognition module. The animal behavior recognition module uses the graphical user interface software to label the target joint points belonging to the object in the labeling data on the image one by one, and then The annotation data is stored in the animal behavior recognition module.
The method of biological monitoring by an automated system based on artificial intelligence according to claim 1, characterized in that, in step 6, the partially abnormal identification results are reviewed by the administrator to evaluate whether they need to be re-labeled and classified to form A new tag data set enables the new tag data set to be continuously input to the machine learning machine to retrain the recognition model to improve the accuracy of the recognition model.
The method of monitoring organisms by an automated system based on artificial intelligence as claimed in claim 1, wherein the machine learning algorithm is cluster analysis (Clustering), support vector machine (SVM) or deep learning (Deep Learning). ) or any combination of the above.
An automated system based on artificial intelligence to monitor biological devices, which is installed on a biological breeding cage. The biological breeding cage is equipped with multiple biological breeding cages. It is characterized in that the automatic system based on artificial intelligence to monitor biological devices includes have:

A track unit, which includes at least one track group, the track group has a first track, the first track is provided on one side of the biological breeding cages;

A driving unit electrically connected to the track unit, the driving unit is provided with at least one driver corresponding to the track group, and the driver drives the track group to operate;

A host unit. The host unit is provided with at least one host corresponding to the first track. The bottom of the host is provided with a first set of connecting plates. The first set of connecting plates is provided with a second set of connecting plates. The second set of connecting plates The board is located in this track group;

A sensing unit is provided with a sensor corresponding to the host. The sensor is electrically connected to the host and is located at one end of the host.
The biological monitoring device of an automated system based on artificial intelligence according to claim 8, characterized in that the host is an edge computing host, including a central processing unit, a memory, a graphics computing unit, a peripheral I/O interface, a wireless Transmission and data storage unit, the sensor includes a camera module, infrared detection, temperature meter, humidity meter, directional microphone, vibration meter and pressure meter or any combination of the above.
The biological monitoring device of an automated system based on artificial intelligence as claimed in claim 8, characterized in that one end of the host computer is connected to an adjustment frame, and the adjustment frame includes a front frame body and a rear frame body, and the front frame body Both sides of the rear frame are connected to the rear frame body by a first fixing piece, and the rear frame body is connected to the main unit by a plurality of second fixing pieces.
The artificial intelligence-based automated system monitoring biological device as claimed in claim 8, wherein the first track is arranged horizontally or vertically.
The artificial intelligence-based automatic system biological monitoring device as claimed in claim 8, further comprising a support unit, the support unit is provided with at least one support frame corresponding to the first track, one end of the support frame is fixed It is provided on the biological breeding cage, and the other end is fixed on the first track. In addition, the support unit also includes a supporting frame, which is erected around the biological breeding cage or is provided on the biological breeding cage. On one side, the support frame is composed of a plurality of support rods, so that the entire support frame is in an inverted U shape or square.
The biological monitoring device of an automated system based on artificial intelligence as claimed in claim 12, characterized in that the track set has a first track, two second tracks and a synchronous round rod, and the second tracks are located on the Both sides of the support frame are respectively fixed to the corresponding support rods. The driver is located on one of the second rails. The first rail is fixed by two fixing brackets and each second rail, and the synchronized round rod is Both ends are respectively fixed on each second track.
The biological monitoring device of an automated system based on artificial intelligence as claimed in claim 8, further comprising a power supply unit, the power supply unit is provided with a power supply module corresponding to the host, and the power supply module provides The power required by the host and the sensor.
The biological monitoring device of an automated system based on artificial intelligence as claimed in claim 8, wherein the first set of connecting plates is provided with a plurality of sliders, and the second set of connecting plates corresponds to the first set of connecting plates. The slide blocks are provided with a plurality of slide grooves, and the slide blocks are fastened to the slide grooves, so that the main unit is placed on the first track and moves on the first track.
The biological monitoring device of an automated system based on artificial intelligence according to claim 8, characterized in that the first set of connecting plates is provided with a plurality of magnetic attracting members, and the second set of connecting plates corresponds to the first set of connecting plates. The magnetic parts of the board are provided with a plurality of magnetic grooves, and the magnetic parts are attracted to the magnetic grooves, so that the main unit is installed on the first track and moves on the first track.
The biological monitoring device of an automated system based on artificial intelligence according to claim 8, characterized in that a pressing member is provided on both sides of the first set of connecting plates, and the second set of connecting plates corresponds to the first set of connecting plates. The two pressing parts of the board are provided with two pressing holes, and the pressing parts are clamped in the pressing holes, so that the main unit is placed on the first track and moves on the first track.
The biological monitoring device of an automated system based on artificial intelligence as claimed in claim 8, wherein the first set of connecting plates is provided with a buckle on both sides, and the second set of connecting plates corresponds to the first set of connecting plates. The two buckle parts of the assembly plate are provided with two buckle holes, and the buckle parts are fastened to the buckle holes, so that the main unit is placed on the first track and on the first track. move.