Family old-age care service method and system based on AI visual identification

Abstract

The invention discloses a family old-age service method and system based on AI visual identification, and belongs to the technical field of AI Internet of Things, and the method comprises the following steps: a monitoring room is provided with a central server, and an old-age service robot is in remote communication with the central server; a human body database is preset in the central server, the central server adjusts a human body action model according to an instruction to generate an action comparison animation, the central server sets a human body state label, and the old-age service robot shoots video information and uploads the video information to the central server; the central server analyzes the human body contour and the face image in the video information and cuts out the human body contour and the face image, the central server searches the closest action comparison animation and the face model, judges the display video information of the group where the searched human body action model or the face model is located, and gives an alarm or gives an early warning prompt; the system has the advantages that the state of the old can be automatically monitored, workers are reminded in time when the old may be in danger, and manpower resources are saved.

Description

Technical Field

[0001] The present invention relates to the field of AI Internet of Things, and in particular to a home elderly care service method and system based on AI visual recognition. Background Art

[0002] At present, home elderly care service refers to a series of services provided by the government and society for the elderly on the basis of the family members' obligation of support and maintenance, aiming to meet the diversified needs of the elderly in terms of daily life care, health management, spiritual comfort, etc. Home elderly care service takes the family as the core and the community as the support, providing socialized services mainly to solve the daily life difficulties for the elderly living at home. The elderly often face the situation of being alone at home or multiple elderly people being at home at the same time. At this time, the elderly can request services by calling the staff or the staff coming to the door regularly. Home elderly care service can reduce the labor cost while meeting the daily needs of the elderly. Elderly care service robots play an important role in the home elderly care service system. Elderly care service robots are devices designed specifically for the elderly and integrated with a variety of intelligent technologies, aiming to assist and improve the quality of life of the elderly. The functions usually include daily care, safety and health care, emotional companionship, cultural and entertainment, remote communication, intelligent monitoring, environmental monitoring, etc.

[0003] The above-mentioned existing technical solutions have the following defects: Although the elderly care service robot has an intelligent monitoring function, it mainly uploads the monitoring information to the control room, and mainly relies on the staff to manually monitor the situation of the elderly. However, the staff cannot view all the monitoring information intensively for a long time, and it is easy to miss something. Summary of the Invention

[0004] In order to automatically monitor the status of the elderly, the present application provides a home elderly care service method and system based on AI visual recognition.

[0005] On the one hand, a home elderly care service method based on AI visual recognition provided by the present application adopts the following technical solutions: A home elderly care service method based on AI visual recognition, comprising the following steps: Set up a central server in the monitoring room, and the elderly care service robot communicates remotely with the central server; The central server presets a human body database, and the human body database pre-stores human body action models and face models; The central server adjusts the human body action model according to the instruction to generate an action comparison animation, and the action comparison animation includes a dynamic human body action model animation and a static human body action model animation; The central server sets human body status tags, associates the action comparison animation and face models in the human body database with the human body tags, and classifies the human body status tags into an alarm group, a warning group and a normal group; The elderly care service robot captures video information and uploads it to the central server. The central server analyzes the human contour and face image in the video information and crops them out. The central server searches for the closest action comparison animation in the human body database according to the human contour, and searches for the closest face model in the human body database according to the face image. The central server determines whether the action comparison animation is a dynamic human body action model animation or a static human body action model animation. If the action comparison animation is a dynamic human body action model animation, then judge the movement direction of the person and control the shooting lens of the elderly care service robot to move synchronously. If the action comparison animation is a static human body action model animation, then perform a count. If the count reaches the set number of times, an early warning prompt is issued. When the next received action comparison animation is a dynamic human body action model animation, the count is cleared. The central server determines the group where the found human body action model or face model is located. If the group is the alarm group, the video information is displayed and an alarm is issued; if the group is the early warning group, an early warning prompt is issued.

[0006] By adopting the above scheme, while the elderly care service robot serves the elderly, it will capture the video of the elderly. The central server automatically analyzes the video to judge the state of the elderly. If danger occurs, an alarm is issued to remind the staff, and the video is displayed for the staff to judge the state of the elderly. If danger may occur, an early warning is issued to remind the staff. At the same time, it will also control the camera of the elderly care service robot to follow the elderly according to the video. Since the elderly rarely move, when the elderly are stationary, the central server will issue an early warning reminder when the elderly maintain an action for a long time, allowing the staff to check the state of the elderly. It is not necessary for the staff to check the state of the elderly for a long time, saving manpower, and can timely remind the staff when the elderly may be in danger, with better safety.

[0007] Preferably, the step "If the action comparison animation is a dynamic human body action model animation, then judge the movement direction of the person and control the shooting lens of the elderly care service robot to move synchronously" further includes: The central server marks the upper body center pivot point and the lower body center pivot point of the human body action model. The upper body center pivot point is located at the chest position of the human body action model, and the lower body center pivot point is located at the waist of the human body action model. When judging the movement direction of the person, determine the upper body center pivot point and the lower body center pivot point in the dynamic human body action model animation, and connect the upper body center pivot point and the lower body center pivot point to obtain an action trend line segment. Judge the movement direction of the person by judging the movement trajectory of the action trend line segment in the dynamic human body action model animation.

[0008] By adopting the above solution, the central server marks multiple fulcrums on the human body motion model. Due to the differences in human body shapes and postures, as well as the limitations of shooting angles, it is possible to make mistakes in judging human body motions directly through contours. With the help of fulcrums, the motions and movement directions of the human body can be judged more accurately.

[0009] Preferably, the following steps are further included: Control the movement of the shooting lens of the elderly care service robot according to the movement trajectory of the action trend line segment in the dynamic human body motion model animation; Set the movement range threshold; When the movement range of only one end point of the action trend line segment is less than the movement range threshold, judge the movement direction of the other end point, and control the shooting lens of the elderly care service robot to continue moving in the above movement direction.

[0010] By adopting the above solution, if the movement range of only one end point of the action trend line segment is less than the movement range threshold, it means that when one fulcrum of the elderly person remains stationary, the other fulcrum moves quickly, usually when the elderly person bends down or makes other such actions. At this time, the lens of the elderly care service robot can move a little more in the deflection direction of the human body to ensure that the video taken can capture the human body contour of the elderly person as completely as possible.

[0011] Preferably, the following steps are further included: The central server marks the upper body central fulcrum, lower body central fulcrum, hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum and foot fulcrum of the human body motion model; After the central server analyzes the human body contour, judge the hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum in the human body contour, and search for the action comparison animation with the same fulcrum according to the judged hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum.

[0012] By adopting the above solution, adding fulcrums at key positions such as the limbs and head can make the central server more accurate in searching for the closest action comparison animation.

[0013] Preferably, the step "if the group is the alarm group, display the video information and issue an alarm; if the group is the early warning group, issue an early warning prompt" further includes: When the central server issues an alarm, perform a stroke processing on the human body contour in the displayed video information, and call the video information of the elderly care service robot within the set time before the displayed video information. If the corresponding action comparison animation in the called video information is the dynamic human body motion model animation, add a direction mark according to the task movement direction, and display the called video information after receiving the display instruction; When the central server issues a warning prompt, it outlines the human body contour in the currently judged video information and displays the video information after receiving the display instruction.

[0014] By adopting the above solution, while the central server issues an alarm, it will also highlight the human body contour in the video, which is convenient for the staff to quickly judge the actions of the elderly. When issuing a warning prompt, it can also prepare the outlined video information in advance to help the staff quickly find the key points.

[0015] On the other hand, a home elderly care service system based on AI visual recognition provided by the present application adopts the following technical solution: A home elderly care service system based on AI visual recognition includes a central server and an elderly care service robot. The elderly care service robot includes a video acquisition module and a lens control module; The video acquisition module acquires video information and uploads it to the central server; The lens control module receives the movement instruction transmitted by the central server and controls the movement of the lens of the elderly care service robot according to the movement instruction; The central server includes an information storage module, an animation generation module, a video processing module, an animation judgment module, and an alarm display module; The information storage module pre-stores human body action models, face models, and human body state labels, associates the action comparison animations and face models in the human body database with the human body labels, and classifies the human body state labels into an alarm group, a warning group, and a normal group; The animation generation module calls the human body action models stored in the information storage module and adjusts the human body action models according to instructions to generate action comparison animations. The action comparison animations include dynamic human body action model animations and static human body action model animations; The video processing module receives the video information uploaded by the elderly care service robot, calls the data stored in the information storage module, analyzes and crops out the human body contour and face image in the video information, finds the closest action comparison animation according to the human body contour, and finds the closest face model according to the face image. The video processing module judges the group where the found human body action model or face model is located. If the group is the alarm group, it sends an alarm signal to the alarm display module. If the group is the warning group, it sends a warning signal to the alarm display module; The animation judgment module calls the action comparison animation of the video processing module to determine whether the action comparison animation is a dynamic human body action model animation or a static human body action model animation. If the action comparison animation is a dynamic human body action model animation, it determines the movement direction of the person and sends a movement instruction to the elderly care service robot. If the action comparison animation is a static human body action model animation, it performs a count. If the count reaches the set number of times, it sends a warning signal to the alarm display module. When the next received action comparison animation is a dynamic human body action model animation, the count is cleared. When the alarm display module receives an alarm signal, it calls the video information of the video processing module to display and issue an alarm. When the alarm display module receives a warning signal, it issues a warning prompt.

[0016] By adopting the above solution, while serving the elderly, the elderly care service robot will take videos of the elderly. The central server automatically analyzes the videos to judge the status of the elderly. If danger occurs, it will issue an alarm to remind the staff and display the videos for the staff to judge the status of the elderly. If danger may occur, it will issue a warning to remind the staff. At the same time, it will also control the camera of the elderly care service robot to follow the movement of the elderly according to the videos. Since the elderly rarely move, when the elderly are stationary, the central server will issue a warning reminder when the elderly maintain a posture for a long time, allowing the staff to check the status of the elderly. It does not require the staff to check the status of the elderly for a long time, saving manpower, and can timely remind the staff when danger may occur to the elderly, with better safety.

[0017] Preferably, the animation judgment module marks the upper body center pivot point and the lower body center pivot point of the human body action model. The upper body center pivot point is located at the chest position of the human body action model, and the lower body center pivot point is located at the waist of the human body action model. When determining the movement direction of the person, it determines the upper body center pivot point and the lower body center pivot point in the dynamic human body action model animation, and connects the upper body center pivot point and the lower body center pivot point to obtain an action trend line segment. It determines the movement direction of the person by judging the movement trajectory of the action trend line segment in the dynamic human body action model animation.

[0018] By adopting the above solution, the central server marks multiple pivot points on the human body action model. Due to the differences in human body shape and posture, as well as the limitation of the shooting angle, it is possible to make mistakes in judging human body actions directly through the contour. Through the pivot points, it is possible to more accurately judge the actions and movement directions of the human body.

[0019] Preferably, the central server further includes a movement prediction module; The information storage module pre-stores a movement range threshold; The movement prediction module calls the dynamic human motion model animation of the animation judgment module, generates a movement instruction according to the movement trajectory of the action trend line segment in the dynamic human motion model animation, and sends the movement instruction to the elderly care service robot. When the movement range of only one endpoint of the action trend line segment is less than the movement range threshold, it judges the movement direction of the other endpoint, generates a delayed movement instruction, and sends the delayed movement instruction to the elderly care service robot; After receiving the delayed movement instruction, the camera control module controls the camera to continue moving according to the instruction.

[0020] By adopting the above solution, if the movement range of only one endpoint of the action trend line segment is less than the movement range threshold, it means that the elderly person quickly moves the other fulcrum while keeping one fulcrum still. Generally, it is when the elderly person bends down or performs other actions. At this time, the camera of the elderly care service robot can move a little more in the deflection direction of the human body to ensure that the video captured can capture the human body contour of the elderly person as completely as possible.

[0021] Preferably, the information storage module marks the upper body center fulcrum, lower body center fulcrum, hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum and foot fulcrum of the human motion model; After the video processing module analyzes the human body contour, it judges the hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum in the human body contour, and searches for the action comparison animation with the same fulcrum according to the judged hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum.

[0022] By adopting the above solution, adding fulcrums at key positions such as the limbs and head can make the central server more accurate in searching for the most similar action comparison animation.

[0023] Preferably, when the alarm display module receives an alarm signal, it performs an outlining process on the human body contour in the displayed video information, and calls the video information of the elderly care service robot within the set time before the displayed video information. If the corresponding action comparison animation in the called video information is a dynamic human motion model animation, it adds a direction mark according to the task movement direction and displays the called video information after receiving the display instruction; when the alarm display module receives a warning message, it performs an outlining process on the human body contour in the currently judged video information and displays the video information after receiving the display instruction.

[0024] By adopting the above solution, when the central server issues an alarm, it will also highlight the human body contour in the video, which is convenient for the staff to quickly judge the actions of the elderly. When issuing a warning prompt, it can also prepare the outlined video information in advance to help the staff quickly find the key points.

[0025] In summary, the present invention has the following beneficial effects: 1. While providing services to the elderly, the elderly care service robot will take videos of the elderly. The central server automatically analyzes the videos to judge the status of the elderly. If danger occurs, it will issue an alarm to remind the staff and display the videos for the staff to judge the status of the elderly. If danger may occur, it will issue a warning to remind the staff. At the same time, it will also control the camera of the elderly care service robot to follow the elderly according to the videos. There is no need for the staff to check the status of the elderly for a long time, which saves manpower and can timely remind the staff when danger may occur to the elderly, with better safety.

[0026] 2. Since the elderly move very little, when the elderly are stationary, the central server will issue a warning reminder when the elderly maintain a single action for a long time, so that the staff can check the status of the elderly. Description of the Drawings

[0027] Figure 1 It is the overall system block diagram of the second embodiment of the present application.

[0028] Figure 2 It is the block diagram of the central server and the elderly care service robot in the second embodiment of the present application.

[0029] Description of the Reference Numerals: 1. Central server; 11. Information storage module; 12. Animation generation module; 13. Video processing module; 14. Animation judgment module; 15. Alarm display module; 16. Movement prediction module; 2. Elderly care service robot; 21. Video acquisition module; 22. Lens control module. Detailed Embodiment

[0030] Embodiment 1. The embodiment of the present application discloses a home elderly care service method based on AI vision recognition, and the specific steps are as follows: S100. Set up a central server 1 in the monitoring room, and the elderly care service robot 2 communicates with the central server 1 remotely. The central server 1 can display information through the display screen in the monitoring room or receive instructions input by the staff.

[0031] S200. The central server 1 presets a human body database, which stores a movement range threshold, a human body motion model, and a face model.

[0032] S201. The central server 1 adjusts the human body motion model according to the instructions to generate a motion comparison animation, which includes a dynamic human body motion model animation and a static human body motion model animation.

[0033] S202. The central server 1 sets human body status tags, associates the motion comparison animation and the face model in the human body database with the human body tags, and classifies the human body status tags into an alarm group, a warning group, and a regular group.

[0034] S300. The elderly care service robot 2 captures video information and uploads it to the central server 1.

[0035] S400. The central server 1 analyzes the human contour and face image in the video information and crops them out.

[0036] S401. The central server 1 searches for the closest action comparison animation in the human body database based on the human contour, and searches for the closest face model in the human body database based on the face image.

[0037] S402. The central server 1 determines whether the action comparison animation is a dynamic human body action model animation or a static human body action model animation.

[0038] S403. If the action comparison animation is a dynamic human body action model animation, then determine the movement direction of the person and control the synchronous movement of the shooting lens of the elderly care service robot 2.

[0039] S404. The central server 1 marks the upper body center pivot point, lower body center pivot point, hand pivot point, elbow pivot point, head pivot point, knee pivot point, and foot pivot point of the human body action model. The upper body center pivot point is located at the chest position of the human body action model, and the lower body center pivot point is located at the waist of the human body action model.

[0040] S405. After the central server 1 analyzes the human contour, it determines the hand pivot point, elbow pivot point, head pivot point, knee pivot point, or foot pivot point in the human contour, and searches for the action comparison animation with the same pivot point according to the determined hand pivot point, elbow pivot point, head pivot point, knee pivot point, or foot pivot point. Adding pivot points at key positions such as the limbs and head can make the central server 1 more accurate in searching for the closest action comparison animation.

[0041] S406. When determining the movement direction of the person, determine the upper body center pivot point and lower body center pivot point in the dynamic human body action model animation, and connect the upper body center pivot point and the lower body center pivot point to obtain an action trend line segment. Determine the movement direction of the person by judging the movement trajectory of the action trend line segment in the dynamic human body action model animation. Control the movement of the shooting lens of the elderly care service robot 2 according to the movement trajectory of the action trend line segment in the dynamic human body action model animation.

[0042] S407. When the moving range of only one endpoint of the movement tendency line segment is less than the movement range threshold, determine the movement direction of the other endpoint, and control the shooting lens of the elderly care service robot 2 to continue moving in the above movement direction. If the moving range of only one endpoint of the movement tendency line segment is less than the movement range threshold, it means that the elderly person moves quickly with one fulcrum stationary while the other fulcrum moves. Generally, the elderly person is performing actions such as bending over. At this time, the lens of the elderly care service robot 2 can move a little more in the direction of the human body deflection to ensure that the video captured can cover as much of the elderly person's body contour as possible.

[0043] S408. If the action comparison animation is a static human body action model animation, perform a count once. If the count reaches the set number of times, issue a warning prompt. When the next received action comparison animation is a dynamic human body action model animation, clear the count.

[0044] S500. The central server 1 determines the group where the found human body action model or face model is located. If the group is the alarm group, display the video information and issue an alarm; if the group is the early warning group, issue a warning prompt.

[0045] S501. When the central server 1 issues an alarm, perform an edge tracing process on the human body contour in the displayed video information, and call the video information of the elderly care service robot 2 within the set time before the displayed video information. If the corresponding action comparison animation in the called video information is a dynamic human body action model animation, add a direction mark according to the task movement direction, and display the called video information after receiving the display instruction.

[0046] S602. When the central server 1 issues a warning prompt, perform an edge tracing process on the human body contour in the currently judged video information, and display the video information after receiving the display instruction. When the central server 1 issues an alarm, it will also highlight the human body contour in the video to facilitate the staff to quickly judge the actions of the elderly. When issuing a warning prompt, it can also prepare the edge-traced video information in advance to help the staff quickly find the key points.

[0047] The implementation principle of a home-based elderly care service method based on AI visual recognition in an embodiment of this application is as follows: While serving the elderly, the elderly care service robot 2 captures videos of the elderly. The central server 1 automatically analyzes the videos to judge the status of the elderly. If a danger occurs, an alarm is sent to remind the staff, and the video is displayed for the staff to judge the status of the elderly. If a potential danger may occur, a warning is sent to remind the staff. At the same time, the camera of the elderly care service robot 2 is controlled to follow the movement of the elderly according to the video. Since the elderly move very little, when the elderly are stationary, the central server 1 will send a warning reminder when the elderly maintain a single posture for a long time, so that the staff can check the status of the elderly. This eliminates the need for the staff to continuously monitor the status of the elderly, saving manpower, and can timely remind the staff when potential danger may befall the elderly, improving safety.

[0048] Embodiment 2. An embodiment of this application discloses a home-based elderly care service system based on AI visual recognition, as Figure 1 and Figure 2 shown, which includes a central server 1 and an elderly care service robot 2. The elderly care service robot 2 includes a video acquisition module 21 and a lens control module 22. The central server 1 includes an information storage module 11, an animation generation module 12, a video processing module 13, an animation judgment module 14, an alarm display module 15, and a movement prediction module 16.

[0049] As Figure 2 shown, the information storage module 11 pre-stores human body movement models, face models, human body status tags, and movement range thresholds, associates the movement comparison animations and face models in the human body database with the human body tags, and classifies the human body status tags into an alarm group, a warning group, and a normal group. The information storage module 11 marks the upper body center pivot point, lower body center pivot point, hand pivot point, elbow pivot point, head pivot point, knee pivot point, and foot pivot point of the human body movement model.

[0050] As Figure 2 shown, the animation generation module 12 calls the human body movement models stored in the information storage module 11, adjusts the human body movement models according to instructions to generate movement comparison animations, and the movement comparison animations include dynamic human body movement model animations and static human body movement model animations.

[0051] As Figure 2As shown, the video acquisition module 21 acquires video information and uploads it to the central server 1. The video processing module 13 receives the video information uploaded by the elderly care service robot 2, calls the data stored in the information storage module 11, analyzes and crops the human contour and facial image in the video information, searches for the closest action comparison animation according to the human contour, and searches for the closest facial model according to the facial image. The video processing module 13 determines the group where the found human action model or facial model is located. If the group is the alarm group, it sends an alarm signal to the alarm display module 15. If the group is the early warning group, it sends an early warning signal to the alarm display module 15. After the video processing module 13 analyzes the human contour, it determines the hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum in the human contour, and searches for the action comparison animation with the same fulcrum according to the determined hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum.

[0052] As Figure 2 shown, the animation judgment module 14 calls the action comparison animation of the video processing module 13 to judge whether the action comparison animation is a dynamic human action model animation or a static human action model animation. If the action comparison animation is a dynamic human action model animation, it judges the movement direction of the person. The animation judgment module 14 marks the upper body center fulcrum and the lower body center fulcrum of the human action model. The upper body center fulcrum is located at the chest position of the human action model, and the lower body center fulcrum is located at the waist of the human action model. When judging the movement direction of the person, it determines the upper body center fulcrum and the lower body center fulcrum in the dynamic human action model animation, and connects the upper body center fulcrum and the lower body center fulcrum to obtain an action trend line segment, and judges the movement direction of the person by judging the movement trajectory of the action trend line segment in the dynamic human action model animation. It sends a movement instruction to the elderly care service robot 2 according to the movement direction of the person. If the action comparison animation is a static human action model animation, it makes a count once. If the count reaches the set number of times, it sends an early warning signal to the alarm display module 15. When the next received action comparison animation is a dynamic human action model animation, the count is cleared. The camera control module 22 receives the movement instruction transmitted by the central server 1 and controls the camera movement of the elderly care service robot 2 according to the movement instruction.

[0053] As Figure 2 shown, the movement prediction module 16 calls the dynamic human action model animation of the animation judgment module 14, generates a movement instruction according to the movement trajectory of the action trend line segment in the dynamic human action model animation, and sends the movement instruction to the elderly care service robot 2. When the movement range of only one endpoint of the action trend line segment is less than the movement range threshold, it judges the movement direction of the other endpoint, generates a delayed movement instruction, and sends the delayed movement instruction to the elderly care service robot 2. After the camera control module 22 receives the delayed movement instruction, it controls the camera to continue moving according to the instruction.

[0054] As Figure 2 shown, when the alarm display module 15 receives an alarm signal, it strokes the human body contour in the displayed video information, and calls the video information of the elderly care service robot 2 within a preset time before the displayed video information. If the corresponding action comparison animation in the called video information is a dynamic human body action model animation, a direction mark is added according to the task movement direction, and the called video information is displayed after receiving the display instruction. When the alarm display module 15 receives a warning signal, it issues a warning prompt, strokes the human body contour in the currently judged video information, and displays the video information after receiving the display instruction.

[0055] The embodiments of this specific implementation manner are all preferred embodiments of the present invention, and do not limit the protection scope of the present invention accordingly. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A family elderly care service method based on AI visual recognition, characterized in that: The following steps are involved: A central server (1) is provided in the monitoring room, and the elderly care service robot (2) remotely communicates with the central server (1); The central server (1) presets a human body database, wherein the human body database pre-stores a human body action model and a face model; The central server (1) adjusts the human motion model according to the instruction to generate a motion comparison animation, wherein the motion comparison animation includes a dynamic human motion model animation and a static human motion model animation; The central server (1) sets a human body state label, associates the action comparison animation and the face model in the human body database with the human body label, and classifies the human body state label into an alarm group, a warning group, and a regular group; The elderly care service robot (2) captures video information and uploads it to the central server (1); The central server (1) analyzes the human body contour and facial image in the video information and cuts them out; The central server (1) searches for the closest action comparison animation in the human body database according to the human body contour, and searches for the closest face model in the human body database according to the face image; The central server (1) determines whether the action comparison animation is a dynamic human action model animation or a static human action model animation; If the action comparison animation is a dynamic human action model animation, the direction of the person's movement is determined, and the elderly care service robot (2) is controlled to move the shooting lens synchronously; If the action comparison animation is a static human action model animation, a count is performed. If the count reaches the set number of times, an early warning prompt is issued. When the next action comparison animation received is a dynamic human action model animation, the count is cleared. The central server (1) determines the group to which the human motion model or face model being searched belongs. If the group is an alarm group, the central server (1) displays the video information and issues an alarm; if the group is an early warning group, an early warning prompt is issued.

2. According to the AI ​​visual recognition-based family elderly care service method and system according to claim 1, it is characterized in that: The step of "if the action comparison animation is a dynamic human action model animation, determining the direction of the person's movement, and controlling the elderly care service robot (2) to move the shooting lens synchronously" also includes: The central server (1) marks the upper body center fulcrum and the lower body center fulcrum of the human motion model, wherein the upper body center fulcrum is located at the chest of the human motion model, and the lower body center fulcrum is located at the waist of the human motion model; When judging the direction of movement of a character, determine the center fulcrum of the upper body and the center fulcrum of the lower body in the dynamic human motion model animation, and connect the center fulcrum of the upper body and the center fulcrum of the lower body to obtain the motion trend line segment. The direction of movement of the character is judged by judging the motion trajectory of the motion trend line segment in the dynamic human motion model animation.

3. According to claim 2, a method and system for family elderly care services based on AI visual recognition is characterized in that: The following steps are also included: Controlling the movement of the shooting lens of the elderly care service robot (2) according to the motion trajectory of the motion trend line segment in the dynamic human motion model animation; Set the moving range threshold; When the movement range of only one endpoint of the motion trend line segment is smaller than the movement range threshold, the movement direction of the other endpoint is determined, and the camera lens of the elderly care service robot (2) is controlled to continue to move in the above movement direction.

4. According to claim 2, a method and system for family elderly care services based on AI visual recognition is characterized in that: The following steps are also included: The central server (1) marks the upper body center fulcrum, the lower body center fulcrum, the hand fulcrum, the elbow fulcrum, the head fulcrum, the knee fulcrum and the foot fulcrum of the human motion model; After analyzing the human body contour, the central server (1) determines the hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum in the human body contour, and searches for action comparison animations with the same fulcrum based on the determined hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum.

5. According to the method and system for family elderly care services based on AI visual recognition according to claim 1, it is characterized in that: The step of "if the group is an alarm group, displaying video information and issuing an alarm; if the group is an early warning group, issuing an early warning prompt" also includes: When the central server (1) issues an alarm, it performs outline processing on the human body contour in the displayed video information, and calls the elderly care service robot (2) to display video information within a set time before the displayed video information. If the corresponding action contrast animation in the called video information is a dynamic human action model animation, a mark indicating the direction is added according to the task movement direction, and the called video information is displayed after receiving the display instruction; When the central server (1) issues a warning prompt, it performs outline processing on the human body contour in the currently judged video information, and displays the video information after receiving a display instruction.

6. A family elderly care service system based on AI visual recognition, characterized by: It comprises a central server (1) and an elderly care service robot (2), wherein the elderly care service robot (2) comprises a video acquisition module (21) and a lens control module (22); The video acquisition module (21) acquires video information and uploads it to the central server (1); The lens control module (22) receives a movement instruction transmitted by the central server (1), and controls the movement of the lens of the elderly care service robot (2) according to the movement instruction; The central server (1) comprises an information storage module (11), an animation generation module (12), a video processing module (13), an animation judgment module (14), and an alarm display module (15); The information storage module (11) pre-stores a human action model, a facial model, and a human state label, associates the action comparison animation and the facial model in the human database with the human state label, and classifies the human state label into an alarm group, a warning group, and a regular group; The animation generation module (12) calls the human action model stored in the information storage module (11), and adjusts the human action model according to the instruction to generate an action comparison animation, wherein the action comparison animation includes a dynamic human action model animation and a static human action model animation; The video processing module (13) receives video information uploaded by the elderly care service robot (2), calls data stored in the information storage module (11), analyzes and cuts out the human body contour and facial image in the video information, searches for the closest action comparison animation according to the human body contour, and searches for the closest facial model according to the facial image. The video processing module (13) determines the group to which the searched human body action model or facial model belongs, and if the group is an alarm group, sends an alarm signal to the alarm display module (15); if the group is an early warning group, sends an early warning signal to the alarm display module (15); The animation judgment module (14) calls the action comparison animation of the video processing module (13) to judge whether the action comparison animation is a dynamic human action model animation or a static human action model animation. If the action comparison animation is a dynamic human action model animation, the motion direction of the character is judged and a movement instruction is sent to the elderly care service robot (2); if the action comparison animation is a static human action model animation, a count is performed, and if the count reaches a set number of times, a warning signal is sent to the alarm display module (15); when the action comparison animation received next time is a dynamic human action model animation, the count is cleared; When the alarm display module (15) receives an alarm signal, it calls the video information display of the video processing module (13) and issues an alarm. When the alarm display module (15) receives an early warning signal, it issues an early warning prompt.

7. The family elderly care service system based on AI visual recognition according to claim 6 is characterized by: The animation judgment module (14) marks the upper body center fulcrum and the lower body center fulcrum of the human action model, wherein the upper body center fulcrum is located at the chest position of the human action model, and the lower body center fulcrum is located at the waist of the human action model; when judging the direction of motion of the character, the upper body center fulcrum and the lower body center fulcrum in the dynamic human action model animation are determined, and the upper body center fulcrum and the lower body center fulcrum are connected to obtain a motion trend line segment, and the motion direction of the character is judged by judging the motion trajectory of the motion trend line segment in the dynamic human action model animation.

8. The family elderly care service system based on AI visual recognition according to claim 7 is characterized by: The central server (1) further comprises a movement prediction module (16); The information storage module (11) pre-stores a moving range threshold; The movement prediction module (16) calls the dynamic human motion model animation of the animation judgment module (14), generates a movement instruction according to the motion trajectory of the motion trend line segment in the dynamic human motion model animation, and sends the movement instruction to the elderly care service robot (2); when the movement range of only one endpoint of the motion trend line segment is less than the movement range threshold, the movement direction of the other endpoint is judged, a delayed movement instruction is generated, and the delayed movement instruction is sent to the elderly care service robot (2); After receiving the delayed movement instruction, the lens control module (22) controls the lens to continue moving according to the instruction.

9. The family elderly care service system based on AI visual recognition according to claim 7 is characterized by: The information storage module (11) marks the upper body center fulcrum, lower body center fulcrum, hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum and foot fulcrum of the human action model; After analyzing the human body contour, the video processing module (13) determines the hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum in the human body contour, and searches for action comparison animations with the same fulcrum based on the determined hand fulcrum, elbow fulcrum, head fulcrum, knee fulcrum or foot fulcrum.

10. The family elderly care service system based on AI visual recognition according to claim 6 is characterized by: When the alarm display module (15) receives an alarm signal, it performs a contour processing on the human body contour in the displayed video information, and calls the video information of the elderly care service robot (2) within a set time before the displayed video information. If the corresponding action contrast animation in the called video information is a dynamic human action model animation, a mark indicating the direction is added according to the task movement direction, and the called video information is displayed after receiving the display instruction; When the alarm display module (15) receives the warning information, it performs outline processing on the human body contour in the currently judged video information, and displays the video information after receiving the display instruction.

Images