An AI vision-based skipping rope abnormal behavior recognition method and system

Abstract

The application discloses a skipping rope abnormal behavior recognition method and system based on AI vision, and relates to the technical field of AI vision recognition.The method comprises the following steps: collecting an initial visual image of a user, extracting human key point information through an AI vision recognition model, completing body type classification, and obtaining body type category parameters; dynamically adjusting the jumping amplitude and jumping height judgment threshold according to the body type category parameters, establishing individualized abnormal judgment criteria; collecting visual images in the skipping process in real time, extracting human key point dynamic data and skipping motion data; analyzing the data based on the individualized abnormal judgment criteria, judging whether there is abnormal behavior, and outputting corresponding warning information.The AI vision processing module in the system contains a refined threshold adjustment unit, which realizes accurate correction of the threshold.The application adjusts the judgment criteria by adapting to the user's body type, solves the problems of low recognition accuracy and high misjudgment rate caused by ignoring body type differences in the prior art, and improves the accuracy and pertinence of abnormal recognition.

Description

Technical Field

[0001] This invention relates to the technical field of AI visual recognition, and more particularly to a method and system for identifying abnormal rope skipping behavior based on AI vision. Background Technology

[0002] Rope skipping, as a simple and efficient fitness exercise, is widely used in daily fitness, sports training, and physical fitness testing. With the development of AI vision technology, rope skipping behavior recognition technology based on AI vision is gradually replacing traditional manual counting and observation methods, realizing automated monitoring of the rope skipping process and early warning of abnormal behavior, and providing users with scientific rope skipping guidance.

[0003] Existing AI vision-based methods for identifying abnormal behavior in rope skipping mostly employ fixed thresholds for jump amplitude and height, ignoring the differences in body type among users. For example, slender users and robust users have different physiological structures, resulting in significant differences in their normal jump amplitude and height. Similarly, short users and standard-sized users exhibit different movement characteristics. Using uniform thresholds for anomaly detection can easily lead to misjudgments of abnormal behavior in users with specific body types (e.g., classifying a slender user's normal small-amplitude jumps as abnormal, or a robust user's abnormal large-amplitude jumps as normal), resulting in low accuracy and failing to meet the personalized monitoring needs of users with different body types.

[0004] Meanwhile, the threshold adjustment logic of the AI ​​vision processing module in the existing technology is relatively crude, and the precise correction process based on body shape parameters is not clearly defined, which makes it difficult to establish personalized judgment standards. In addition, some recognition methods do not have a sound abnormality calibration mechanism and graded early warning system, which can easily lead to misjudgment due to sudden interference or accidental movement deviation. Furthermore, they cannot provide targeted early warnings based on the severity of the abnormality, making it difficult to effectively ensure the safety of users jumping rope and hindering the actual promotion and application of the technology. Summary of the Invention

[0005] The present invention aims to overcome the shortcomings of the prior art, and the technical solution adopted by the present invention is as follows: In a first aspect of the present invention, a method for identifying abnormal rope skipping behavior based on AI vision is provided, comprising the following steps: S1. Collect the initial visual image of the rope skipping user, extract the key point information of the user's human body through the AI ​​visual recognition model, and complete the user's body type classification based on the key point information to obtain the body type category parameter. S2. Based on the body type category parameters, dynamically adjust the jump amplitude judgment threshold and jump height judgment threshold to establish a personalized anomaly judgment standard that matches the user's body type; S3. Real-time acquisition of continuous visual images of the user during rope skipping, and real-time extraction of dynamic data of key human body points and rope skipping motion data through the AI ​​visual recognition model; S4. Based on the personalized anomaly judgment criteria, analyze the dynamic data of the human body key points and the rope skipping exercise data to determine whether the user has abnormal rope skipping behavior. If so, output the abnormal behavior type and warning information.

[0006] Preferably, the key point information includes two-dimensional coordinates of the head, shoulders, waist, hips, knees, and ankles; the body type classification is based on the shoulder-to-hip ratio, height-to-leg-length ratio, and body fat percentage correlation features, classifying user body types into slender, standard, stocky, and short types, and the body type category parameters include the feature quantification values ​​of the corresponding body type.

[0007] Preferably, the specific process of dynamically adjusting the jump amplitude judgment threshold and the jump height judgment threshold is as follows: S21. Preset the baseline jump amplitude threshold and baseline jump height threshold for each body type. The baseline jump amplitude threshold is defined as the range of angle changes between the hip and knee joints of the human body during rope skipping. The baseline jump height threshold is defined as the range of vertical distances from the soles of the feet to the ground of the human body during rope skipping. S22. Based on the feature quantification values ​​in the body type category parameters, the baseline jump amplitude threshold and baseline jump height threshold are corrected by a preset adjustment algorithm to obtain a personalized judgment threshold that matches the current user's body type; wherein, the jump amplitude threshold and jump height threshold for slender users are lowered by 5%-15%, the jump amplitude threshold and jump height threshold for stocky users are raised by 5%-15%, and the jump amplitude threshold for short users is raised by 3%-10% and the jump height threshold is lowered by 3%-10%.

[0008] More preferably, the correction process of the preset adjustment algorithm is as follows: First, the preset baseline jump amplitude threshold and baseline jump height threshold for each body type category are retrieved from the data storage module. Second, the feature quantization value in the body type category parameter is extracted and input into a preset adjustment algorithm. The algorithm determines the threshold correction coefficient based on the deviation ratio between the feature quantization value and the corresponding body type baseline value. Finally, the baseline jump amplitude threshold and baseline jump height threshold are precisely corrected based on the correction coefficient, and a personalized anomaly judgment standard matching the current user's body type is generated after correction.

[0009] Preferably, the dynamic data of the human body key points includes displacement, velocity, and acceleration data of each human body key point, as well as angular change data between adjacent key points; the rope skipping motion data includes the swing frequency, swing amplitude, and relative position data of the rope and the human body.

[0010] Preferably, the abnormal rope skipping behaviors include abnormal jump amplitude, abnormal jump height, abnormal rope swinging motion, abnormal landing posture, and abnormal rope tripping; wherein: Abnormal jump amplitude refers to the change in the angle between the human hip and knee joints exceeding the personalized jump amplitude judgment threshold. An abnormal jump height refers to a vertical distance from the sole of the foot to the ground that exceeds the personalized jump height judgment threshold. Abnormal rope swinging movements include excessive involvement of the upper arm in rope swinging, stiff wrists leading to uneven rope swinging amplitude, and mismatch between rope swinging frequency and jumping frequency. Abnormal landing postures include landing on the entire foot or heel first, loss of balance upon landing, and excessive extension of the knee joint; A skipping rope tripping abnormality refers to a situation where the skipping rope comes into contact with the feet of the person during the skipping rope exercise without completing the normal jumping action.

[0011] Preferably, the AI ​​visual recognition model includes a human body detection module, a key point extraction module, and a body shape classification module. The human body detection module uses a Mask RCNN instance segmentation neural network to achieve accurate segmentation of the user's human body region. The key point extraction module uses the OpenPose algorithm to extract the coordinates of 25 key points of the human body. The body shape classification module uses a multimodal large model to complete body shape classification based on the contour line features of the human body's key points, avoiding the interference of clothing and hairstyle factors that affect the accuracy of judgment.

[0012] As a preferred option, S4 also includes an abnormal behavior calibration step: when a suspected abnormal behavior is detected, visual images of that moment and a preset number of frames before and after are extracted, and compared and analyzed in conjunction with the user's historical jump rope data to eliminate misjudgments caused by sudden interference or accidental deviations in movement. If the abnormal behavior is confirmed, different levels of warning information are output according to the severity of the abnormality.

[0013] Preferably, the severity of the abnormality is divided into three levels, corresponding to different warning messages: Level 1 is a minor abnormality, which is indicated by a text warning displayed on an external screen, only informing the user of the type of abnormal behavior and not affecting the rope skipping process; Level 2 is a moderate abnormality, which is indicated by a text warning displayed on an external screen and an audio warning, reminding the user to adjust their movements in time to avoid the abnormality from escalating; Level 3 is a severe abnormality, which is indicated by a text warning displayed on an external screen, an audio warning, and a light warning, while also triggering a rope skipping exercise pause prompt to prevent physical injury caused by abnormal movements.

[0014] In another aspect of the present invention, an AI vision-based rope skipping abnormal behavior recognition system is also provided, applied to the aforementioned AI vision-based rope skipping abnormal behavior recognition method, comprising:

[0015] The image acquisition module is used to acquire the user's initial visual image and continuous visual images during the rope skipping process. The image acquisition module includes at least one high-definition camera, which can acquire the user's front, side and rear views to ensure the integrity of the extraction of key points of the human body. The AI ​​vision processing module is connected to the image acquisition module and is used to extract key human body information, classify user body types, dynamically adjust the threshold for judging jump amplitude and jump height, and analyze motion data in real time during the rope skipping process. An anomaly detection module, connected to the AI ​​vision processing module, is used to determine whether a user has abnormal rope skipping behavior based on personalized anomaly detection criteria, and to determine the type of abnormal behavior. The early warning output module is connected to the anomaly judgment module and is used to output the type of abnormal behavior and early warning information, including sound warning, text prompt, and light warning. The data storage module, connected to the AI ​​vision processing module and the anomaly detection module, is used to store user body shape data, personalized judgment thresholds, rope skipping exercise data, abnormal behavior records, and historical comparison data.

[0016] Compared with the prior art, the present invention has the following significant advantages: 1. This invention extracts key human body information and classifies body types using an AI visual recognition model. It dynamically adjusts the threshold for judging jump amplitude and jump height based on body type parameters, and establishes a personalized abnormal judgment standard that matches the user's body type. This effectively solves the technical problems of low recognition accuracy and high false judgment rate caused by ignoring body type differences in existing technologies. It significantly improves the accuracy and pertinence of abnormal rope skipping behavior recognition and adapts to the personalized monitoring needs of users with different body types, such as slender, standard, stocky, and short.

[0017] 2. This invention refines the working process of the threshold adjustment unit in the AI ​​vision processing module, clarifies the complete process of benchmark threshold calling, correction coefficient determination, threshold precision correction and standard synchronization, and combines body shape feature quantification value and preset adjustment algorithm to achieve precise correction of judgment threshold, making personalized anomaly judgment standard clearly implementable and facilitating the application of the technology.

[0018] 3. This invention establishes a comprehensive abnormal behavior calibration process. By extracting visual images of suspected abnormal moments and comparing them with the user's historical jump rope data, it eliminates misjudgments caused by sudden interference and accidental movement deviations, further improving the reliability of the recognition results. At the same time, it establishes a correspondence between three levels of abnormal severity and warning information, and outputs warnings in different forms such as text, sound, and light according to the abnormality level, and even triggers exercise pause prompts, effectively ensuring the user's jump rope safety and avoiding physical injury caused by abnormal movements.

[0019] 4. The system structure of this invention is complete. The image acquisition module can acquire images from multiple perspectives of the user to ensure the integrity of human body key point extraction. The AI ​​vision processing module has clear division of labor and works in concert to realize integrated processing of human body detection, key point extraction, body shape classification and threshold adjustment. The interactive module realizes a closed loop of user feedback and can further optimize recognition accuracy through manual calibration, improve user experience and has broad application prospects.

[0020] 5. This invention uses AI vision algorithms combined with multimodal large models to complete body shape classification, effectively avoiding the influence of interference factors such as clothing and hairstyle, improving the accuracy of human body key point extraction and body shape classification, providing reliable data support for personalized threshold adjustment and anomaly recognition, and reducing the difficulty and cost of technical implementation. Attached Figure Description

[0021] Figure 1 This is a flowchart of a method for identifying abnormal rope skipping behavior based on AI vision in a specific embodiment of the present invention; Figure 2 This is a flowchart illustrating the overall process of dynamically adjusting the jump amplitude and jump height judgment thresholds in a specific embodiment of the present invention. Figure 3 This is a framework diagram of an AI vision-based rope skipping abnormal behavior recognition system in a specific embodiment of the present invention; Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] refer to Figure 1 As shown, this embodiment provides a method for identifying abnormal skipping rope behavior based on AI vision. The specific steps are as follows, and a detailed explanation is provided for different age groups and body types, taking into account the home fitness scenario: S1: Acquire the initial visual image of the rope skipping user, extract key human body point information using an AI visual recognition model, and classify the user's body type based on the key human body point information to obtain body type category parameters, specifically including: First, an image acquisition environment was set up by placing one high-definition camera each in front of, to the left of, and behind the user's jump rope area to ensure that the cameras could fully capture the user's full-body movements without obstruction. The user stood in the center of the jump rope area, maintaining a natural standing posture. The cameras simultaneously acquired three sets of initial visual images, including a front view, a side view, and a rear view, and then transmitted the acquired image data to the AI ​​vision processing module.

[0024] The human detection unit in the AI ​​vision processing module uses a Mask RCNN instance segmentation neural network to process the initial visual image, accurately segmenting the user's human body area and eliminating interference factors such as background and jump rope equipment. The key point extraction unit uses the OpenPose algorithm to extract the three-dimensional coordinate data of 25 key points of the human body, focusing on extracting the coordinate information of the head, left shoulder, right shoulder, waist, left hip, right hip, left knee, right knee, left ankle, and right ankle. The shoulder-to-hip ratio (the ratio of the distance from the midpoint of the line connecting the left and right shoulders to the midpoint of the line connecting the left and right hips) and the height-to-leg-length ratio (the ratio of the distance from the top of the head to the sole of the foot to the distance from the midpoint of the hip to the sole of the foot) are calculated as quantitative values ​​of body shape features.

[0025] The body type classification unit employs a multimodal large model. Extracted quantified values ​​of the shoulder-to-hip ratio and height-to-leg-length ratio are input into the model. Combining the contour features of the human body's three-view drawings, and referring to preset body type classification standards, the model completes the classification: a shoulder-to-hip ratio of 0.8-0.9 and a height-to-leg-length ratio of 1.5-1.6 is classified as slender; 0.9-1.0 and 1.4-1.5 as standard; 1.0-1.1 and 1.3-1.4 as stocky; and 0.9-1.0 and 1.2-1.3 as short. After classification, body type parameters are output, including the body type (slender, standard, stocky, or short) and the corresponding quantified values ​​of the shoulder-to-hip ratio and height-to-leg-length ratio. These body type parameters are then synchronized to the threshold adjustment unit and data storage module.

[0026] S2: Based on the body type category parameters, dynamically adjust the jump amplitude judgment threshold and jump height judgment threshold to establish a personalized anomaly judgment standard that matches the user's body type, specifically including: The threshold adjustment unit first calls the preset baseline jump amplitude threshold and baseline jump height threshold corresponding to each body type in the data storage module: among them, the baseline jump amplitude threshold (range of hip and knee joint angle change) for standard users is [30°, 60°], and the baseline jump height threshold (range of vertical distance from the sole of the foot to the ground) is [5cm, 15cm]; the baseline thresholds for slender, stocky, and short users are the same as those for standard users, and are subsequently adjusted by correction coefficients.

[0027] Please see Figure 2The specific process of dynamically adjusting the jump amplitude judgment threshold and the jump height judgment threshold is as follows: S21. Preset the baseline jump amplitude threshold and baseline jump height threshold for each body type. The baseline jump amplitude threshold is defined as the range of angle changes between the hip and knee joints of the human body during rope skipping. The baseline jump height threshold is defined as the range of vertical distances from the soles of the feet to the ground of the human body during rope skipping. S22. Based on the feature quantification values ​​in the body type category parameters, the baseline jump amplitude threshold and baseline jump height threshold are corrected by a preset adjustment algorithm to obtain a personalized judgment threshold that matches the current user's body type; wherein, the jump amplitude threshold and jump height threshold for slender users are lowered by 5%-15%, the jump amplitude threshold and jump height threshold for stocky users are raised by 5%-15%, and the jump amplitude threshold for short users is raised by 3%-10% and the jump height threshold is lowered by 3%-10%.

[0028] Subsequently, the threshold adjustment unit inputs the feature quantization values ​​from the body type category parameters into a preset adjustment algorithm (using a linear correction algorithm). Based on the deviation ratio between the feature quantization values ​​and the corresponding body type benchmark values, the threshold correction coefficient is determined: if the user is slender, and both the shoulder-to-hip ratio and height-to-leg-length ratio are higher than the standard body type benchmark values, the correction coefficient is 0.9; if the user is stocky, and the shoulder-to-hip ratio is higher than the standard body type benchmark value, and the height-to-leg-length ratio is lower than the standard body type benchmark value, the correction coefficient is 1.1; if the user is short, and the shoulder-to-hip ratio is the same as the standard body type, and the height-to-leg-length ratio is lower than the standard body type benchmark value, the jump amplitude threshold correction coefficient is 1.05, and the jump height threshold correction coefficient is 0.95.

[0029] Finally, the baseline thresholds are precisely adjusted based on the determined correction coefficients: for example, the personalized jump amplitude threshold for slender users is 30°×0.9=27° to 60°×0.9=54°, and the personalized jump height threshold is 5cm×0.9=4.5cm to 15cm×0.9=13.5cm; the personalized jump amplitude threshold for robust users is 33° to 66°, and the personalized jump height threshold is 5.5cm to 16.5cm; the personalized jump amplitude threshold for short users is 31.5° to 63°, and the personalized jump height threshold is 4.75cm to 14.25cm. After the adjustment, personalized anomaly judgment criteria are generated and synchronized to the anomaly judgment module and data storage module for subsequent abnormal behavior identification.

[0030] S3: Real-time acquisition of continuous visual images of the user during rope skipping, and extraction of dynamic data of key human body points and rope skipping motion data in real time through the AI ​​visual recognition model, specifically including: After the user starts jumping rope, three high-definition cameras capture continuous visual images of the user's jump rope process in real time, maintaining a constant frame rate. Each frame is transmitted to the AI ​​vision processing module. The human detection unit and key point extraction unit of the AI ​​vision processing module work continuously to extract the three-dimensional coordinate dynamic data of 25 key points of the human body in each frame, including the displacement, velocity, and acceleration data of each key point, as well as the angle change data of the hip and knee joints, and the knee and ankle joints. At the same time, the jump rope motion data is extracted through image recognition algorithms, including the swing frequency of the jump rope (unit: times / minute), the swing amplitude (maximum distance between the two ends of the jump rope), and the relative position data of the jump rope and the human feet, which are updated and stored in real time to the data storage module.

[0031] Based on the personalized anomaly judgment criteria, S4 analyzes the dynamic data of key human body points and rope skipping data to determine whether the user exhibits abnormal rope skipping behavior. If so, it outputs the abnormal behavior type and warning information, specifically including: The anomaly detection module receives real-time dynamic data of key human body points and rope skipping motion data transmitted by the AI ​​vision processing module, and combines this data with personalized anomaly detection criteria to perform real-time analysis and determine whether the user is exhibiting abnormal rope skipping behavior. If a slender user's hip and knee angle changes by 25° (below the lower limit of the personalized jump amplitude threshold of 27° for slender users), it is considered an abnormal jump amplitude; if a stocky user's foot leaves the ground at a vertical distance of 17cm (above the upper limit of the personalized jump height threshold of 16.5cm for stocky users), it is considered an abnormal jump height; if the upper arm swing angle exceeds 45° when swinging the rope, it is considered an abnormal rope swinging motion; if the heel touches the ground first upon landing and the body's center of gravity shifts by more than 10cm, it is considered an abnormal landing posture; if the jump rope contacts the foot and no normal jumping motion is detected within the next 0.5 seconds, it is considered an abnormal jump rope tripping.

[0032] When suspected abnormal behavior is detected, an abnormal behavior calibration step is initiated: visual images of the suspected abnormal moment and the five frames before and after it are extracted, and the user's historical jump rope data from the data storage module is retrieved for comparative analysis. If the comparison reveals that the suspected abnormal behavior is merely an occasional movement deviation, such as a slightly smaller jump amplitude that subsequently returns to normal, it is determined to be a false alarm, and no warning information is output. If the comparison confirms a persistent abnormality, such as abnormal jump height detected in n consecutive frames, then abnormal behavior is confirmed, and a warning information of the corresponding level is output according to the severity of the abnormality. Level 1 Anomaly (Minor Anomaly): Such as a single jump amplitude slightly below the threshold, or slight wrist stiffness causing uneven rope swing amplitude, a text prompt warning will be output, and the connected display screen will show "Jump amplitude is too small, please adjust accordingly", which will not affect the user's rope skipping process; Level 2 Anomaly (Moderate Anomaly): If abnormal jump height or excessive upper arm involvement in rope swinging is detected in two consecutive frames, a text prompt and sound warning will be output, the anomaly type will be displayed on the screen, and a prompt tone will be emitted by the speaker saying "Please note, abnormal movement, please adjust in time" to remind the user to adjust the movement; Level 3 Anomaly (Severe Anomaly): If abnormal landing posture or tripping while jumping rope is detected for 3 consecutive frames, text prompts, sound warnings and light alerts will be output. The anomaly type and risk warning will be displayed on the screen, the speaker will continuously emit a warning sound, the warning lights around the jump rope area will flash red, and the screen will display a prompt "It is recommended to pause jumping rope and adjust your movements before continuing", triggering a sports pause prompt to prevent users from causing ankle and knee injuries due to abnormal movements.

[0033] Please see Figure 3 As shown, in a second aspect of the present invention, an AI vision-based rope skipping abnormal behavior recognition system is proposed, applied to the aforementioned AI vision-based rope skipping abnormal behavior recognition method, comprising:

[0034] The image acquisition module is used to acquire the user's initial visual image and continuous visual images during the rope skipping process. The image acquisition module includes at least one high-definition camera, which can acquire the user's front, side and rear views to ensure the integrity of the extraction of key points of the human body. The AI ​​vision processing module is connected to the image acquisition module and is used to extract key human body information, classify user body types, dynamically adjust the threshold for judging jump amplitude and jump height, and analyze motion data in real time during the rope skipping process. An anomaly detection module, connected to the AI ​​vision processing module, is used to determine whether a user has abnormal rope skipping behavior based on personalized anomaly detection criteria, and to determine the type of abnormal behavior. The early warning output module is connected to the anomaly judgment module and is used to output the type of abnormal behavior and early warning information, including sound warning, text prompt, and light warning. The data storage module, connected to the AI ​​vision processing module and the anomaly detection module, is used to store user body shape data, personalized judgment thresholds, rope skipping exercise data, abnormal behavior records, and historical comparison data.

[0035] The system also includes an interactive module, connected to the AI ​​vision processing module and data storage module. This module includes a touchscreen display and physical buttons for user operation and querying. Users can view their body shape data, personalized anomaly detection thresholds, jump rope exercise records, and abnormal behavior analysis reports through the interactive module. When users believe the body shape classification results or anomaly detection criteria are inaccurate, they can manually calibrate these results and thresholds through the interactive module. Calibration data is synchronized to the AI ​​vision processing module and data storage module, creating a closed-loop user feedback system and further optimizing the system's recognition accuracy. Furthermore, the interactive module allows users to set parameters such as warning volume and warning light brightness, enhancing the user experience.

[0036] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for identifying abnormal rope skipping behavior based on AI vision, characterized in that, Includes the following steps: S1. Collect the initial visual image of the rope skipping user, extract the key point information of the user's human body through the AI ​​visual recognition model, and complete the user's body type classification based on the key point information to obtain the body type category parameter. S2. Based on the body type category parameters, dynamically adjust the jump amplitude judgment threshold and jump height judgment threshold to establish a personalized anomaly judgment standard that matches the user's body type; S3. Real-time acquisition of continuous visual images of the user during rope skipping, and real-time extraction of dynamic data of key human body points and rope skipping motion data through the AI ​​visual recognition model; S4. Based on the personalized anomaly judgment criteria, analyze the dynamic data of the human body key points and the rope skipping exercise data to determine whether the user has abnormal rope skipping behavior. If so, output the abnormal behavior type and warning information.

2. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, The key point information includes two-dimensional coordinates of the head, shoulders, waist, hips, knees, and ankles; the body type classification is based on the shoulder-to-hip ratio, height-to-leg-length ratio, and body fat percentage correlation features, classifying users' body types into slender, standard, stocky, and short types, and the body type category parameters include the feature quantification values ​​of the corresponding body type.

3. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, The specific process for dynamically adjusting the jump amplitude judgment threshold and the jump height judgment threshold is as follows: S21. Preset the baseline jump amplitude threshold and baseline jump height threshold for each body type. The baseline jump amplitude threshold is defined as the range of angle changes between the hip and knee joints of the human body during rope skipping. The baseline jump height threshold is defined as the range of vertical distances from the soles of the feet to the ground of the human body during rope skipping. S22. Based on the feature quantification values ​​in the body type category parameters, the baseline jump amplitude threshold and baseline jump height threshold are corrected by a preset adjustment algorithm to obtain a personalized judgment threshold that matches the current user's body type; wherein, the jump amplitude threshold and jump height threshold for slender users are lowered by 5%-15%, the jump amplitude threshold and jump height threshold for stocky users are raised by 5%-15%, and the jump amplitude threshold for short users is raised by 3%-10% and the jump height threshold is lowered by 3%-10%.

4. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 3, characterized in that, The correction process of the preset adjustment algorithm is as follows: First, the preset baseline jump amplitude threshold and baseline jump height threshold for each body type category are called from the data storage module; second, the feature quantization value in the body type category parameter is extracted, and the feature quantization value is input into the preset adjustment algorithm. The algorithm determines the threshold correction coefficient based on the deviation ratio between the feature quantization value and the corresponding body type baseline value. Finally, based on the correction coefficient, the baseline jump amplitude threshold and the baseline jump height threshold are precisely corrected, and a personalized anomaly judgment standard matching the current user's body shape is generated after correction.

5. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, The dynamic data of key human body points includes displacement, velocity, and acceleration data of each key human body point, as well as angular change data between adjacent key points; the rope skipping motion data includes the swing frequency, swing amplitude, and relative position data of the rope and the human body.

6. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, The abnormal rope skipping behaviors include abnormal jump amplitude, abnormal jump height, abnormal rope swinging motion, abnormal landing posture, and abnormal tripping while skipping rope; among which: Abnormal jump amplitude refers to the change in the angle between the human hip and knee joints exceeding the personalized jump amplitude judgment threshold. An abnormal jump height refers to a vertical distance from the sole of the foot to the ground that exceeds the personalized jump height judgment threshold. Abnormal rope swinging movements include excessive involvement of the upper arm in rope swinging, stiff wrists leading to uneven rope swinging amplitude, and mismatch between rope swinging frequency and jumping frequency. Abnormal landing postures include landing on the entire foot or heel first, loss of balance upon landing, and excessive extension of the knee joint; A skipping rope tripping abnormality refers to a situation where the skipping rope comes into contact with the feet of the person during the skipping rope exercise without completing the normal jumping action.

7. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, The AI ​​visual recognition model includes a human body detection module, a key point extraction module, and a body shape classification module. The human body detection module uses a Mask RCNN instance segmentation neural network to achieve accurate segmentation of the user's human body region. The key point extraction module uses the OpenPose algorithm to extract the coordinates of 25 key points of the human body. The body shape classification module uses a multimodal large model to complete body shape classification based on the contour line features of the human body's key points, avoiding the interference of clothing and hairstyle factors that affect the accuracy of judgment.

8. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, S4 also includes an abnormal behavior calibration step: when a suspected abnormal behavior is detected, visual images of that moment and a preset number of frames before and after are extracted and compared with the user's historical jump rope data to eliminate misjudgments caused by sudden interference or accidental deviations in movement. If the abnormal behavior is confirmed, different levels of warning information are output according to the severity of the abnormality.

9. The method for identifying abnormal rope skipping behavior based on AI vision according to claim 1, characterized in that, The severity of the abnormality is divided into three levels, corresponding to different warning messages: Level 1 is a minor abnormality, which is indicated by a text warning displayed on an external screen, only informing the user of the type of abnormal behavior and not affecting the rope skipping process; Level 2 is a moderate abnormality, which is indicated by a text warning displayed on an external screen and an audio warning, reminding the user to adjust their movements in time to avoid the abnormality from escalating. Level 3 abnormality is a serious abnormality. It will display corresponding text prompts, sound warnings and light warnings on the external display screen, and trigger a stop prompt for rope skipping to prevent physical injury caused by abnormal movements.

10. A jump rope abnormality behavior recognition system based on AI vision, used to implement the method of any one of claims 1-9, characterized in that, include: The image acquisition module is used to acquire the user's initial visual image and continuous visual images during the rope skipping process. The image acquisition module includes at least one high-definition camera, which can acquire the user's front, side and rear views to ensure the integrity of the extraction of key points of the human body. The AI ​​vision processing module is connected to the image acquisition module and is used to extract key human body information, classify user body types, dynamically adjust the threshold for judging jump amplitude and jump height, and analyze motion data in real time during the rope skipping process. An anomaly detection module, connected to the AI ​​vision processing module, is used to determine whether a user has abnormal rope skipping behavior based on personalized anomaly detection criteria, and to determine the type of abnormal behavior. The early warning output module is connected to the anomaly judgment module and is used to output the type of abnormal behavior and early warning information, including sound warning, text prompt, and light warning. The data storage module, connected to the AI ​​vision processing module and the anomaly detection module, is used to store user body shape data, personalized judgment thresholds, rope skipping exercise data, abnormal behavior records, and historical comparison data.

Images