Athletic injury real-time early warning and health optimization system based on multi-mode infrared thermal imaging and three-dimensional skeleton posture fusion

Abstract

The invention relates to a sports injury real-time early warning and health optimization system based on multi-mode infrared thermal imaging and three-dimensional skeleton posture fusion, and belongs to the field of sports health management. The problems of motion analysis lagging, damage identification missing and cross-scene data splitting of a traditional monitoring means are solved. According to the technical scheme, the system comprises a multi-modal motion data sensing module which synchronously collects human thermodynamic distribution, three-dimensional skeleton postures and surface electromyogram signals through an infrared thermal imaging camera, an RGB-D camera and a bioelectric sensor; an AI driving training optimization engine is used for generating an adjustment instruction in real time based on a muscle thermal gradient and joint trajectory deviation degree dual-threshold model; and the full-scene intelligent management platform executes millisecond damage early warning, metabolic optimization and school-family health data closed-loop management. The technical effects are that the exercise safety guarantee is fundamentally improved, the training is scientifically optimized, the metabolism management accuracy is broken through, and the teaching resource allocation is intensive.

Description

Technical Field

[0001] The present invention belongs to the field of sports health management and relates to a real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture. Background Art

[0002] In the fields of physical education and health management, traditional sports monitoring methods mainly rely on wearable devices or optical cameras. These technologies can only obtain single-modality data such as heart rate and two-dimensional motion images, and are unable to simultaneously collect human thermodynamic characteristics, three-dimensional skeletal motion trajectories, and bioelectric signals. Due to the lack of fusion analysis of multi-dimensional data, the system has three major technical bottlenecks:

[0003] (1) Lag in analysis of action norms

[0004] Traditional methods rely on offline video playback for movement correction and cannot provide real-time warning within 100 milliseconds when the joint trajectory deviates from the standard posture. For example, when the deviation of the knee joint movement trajectory exceeds ten percent, the system cannot intervene immediately.

[0005] (2) Lack of identification of sports injury risks

[0006] Existing technologies have difficulty quantifying muscle fatigue, particularly in detecting abnormal thermodynamic distributions within deep muscle groups. For example, a temperature gradient exceeding 2 degrees Celsius per square centimeter over a local muscle region can indicate a risk of muscle strain, but conventional optical devices are unable to detect these thermodynamic signatures.

[0007] (3) Fragmentation of cross-scenario health management

[0008] Data from school training tests, home exercise, and medical examinations are fragmented and independent, making it impossible to construct a dynamically linked student health profile. This results in a lack of consistent data support for key health indicators such as body mass index trend predictions and post-exercise calorie expenditure calculations, leading to errors exceeding 5% in nutritional supplement plans.

[0009] Current technological advancements attempt to optimize motion recognition accuracy through deep learning algorithms, but remain limited by a single data modality. Optical cameras are susceptible to interference from ambient lighting and clothing, making them unable to penetrate the body surface to capture muscle thermodynamics. Wearable devices, on the other hand, limit freedom of movement due to their contact-based measurement methods. Therefore, there is an urgent need to overcome the three major technical barriers of multimodal perception fusion, real-time injury warning, and cross-scenario data closure to promote the intelligent transformation of sports and health management. Summary of the Invention

[0010] In view of this, the object of the present invention is to provide a real-time warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture.

[0011] In order to achieve the above object, the present invention provides the following technical solutions:

[0012] A real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture, including:

[0013] A multimodal motion data perception module configured to simultaneously collect human infrared thermal imaging data, three-dimensional skeletal posture data, and surface electromyography (sEMG) signals;

[0014] An AI-driven training optimization engine, in communication with the multimodal motion data perception module, for generating real-time motion instructions and dynamic training plans based on multimodal data fusion analysis;

[0015] The full-scenario intelligent management platform is connected to the AI-driven training optimization engine to perform injury warning, metabolic optimization and cross-scenario closed-loop management of health data.

[0016] Furthermore, the multimodal motion data perception module includes:

[0017] Infrared thermal imaging unit, using an infrared sensor with a resolution of no less than 160×120 and a temperature measurement range of -40°C to +550°C;

[0018] A 3D skeleton reconstruction unit, including an RGB-D camera and a Time of Flight (TOF) sensor, is configured to extract at least 18 key points of the human body and achieve 3D reconstruction of the skeleton posture with a positioning accuracy of ≤±2mm;

[0019] The bioelectric signal acquisition unit is configured to synchronously acquire surface electromyography (sEMG) signals.

[0020] Furthermore, the AI-driven training optimization engine performs the following operations:

[0021] (a) Establish a dual-threshold judgment model for muscle thermal gradient and joint trajectory deviation:

[0022] When the temperature gradient in the target area is detected Or when the joint trajectory deviation δ>10%, a real-time adjustment instruction is generated;

[0023] in, T is the muscle surface temperature, s is the spatial coordinate;

[0024] P is the joint coordinate;

[0025] (b) Dynamically generate a phased training plan based on the Improved Adaptive Genetic Algorithm Backpropagation (IAGABP).

[0026] Furthermore, the full-scenario intelligent management platform includes:

[0027] The damage warning submodule is configured to trigger an audible and visual alarm or forced load reduction operation within 100ms;

[0028] Metabolic optimization submodule calculates calorie consumption value E based on post-exercise thermal imaging data cal :

[0029] E cal =k·∫ A ΔT(x,y)dA+b·MET

[0030] Wherein, ΔT is the temperature difference distribution on the body surface, A is the effective heat dissipation area, MET is the metabolic equivalent, k and b are calibration coefficients;

[0031] The data closed-loop sub-module integrates school training, home exercise and physical examination data to build a dynamic physical health file.

[0032] Furthermore, the metabolic optimization submodule is based on the calorie consumption value E cal Recommended protein intake M protein :

[0033] M protein =α·E cal +β

[0034] Here, α=0.05g / kcal, and β is a correction term based on the user's physique.

[0035] Furthermore, the infrared thermal imaging unit and the three-dimensional bone reconstruction unit adopt a hardware synchronization trigger mechanism, and the spatiotemporal alignment error is ≤1ms.

[0036] Furthermore, the dual-threshold judgment model is implemented through the DeepSeek-Pro motion optimization large model, which adopts the Transformer architecture and has a parameter volume of ≥120 million.

[0037] Furthermore, the real-time motion instructions are output through augmented reality AR glasses or bone conduction headphones, and the instruction content includes a load adjustment amplitude of ±15%.

[0038] A system-based sports injury early warning and health optimization method, comprising the steps of:

[0039] S1: Synchronously collects the user's thermodynamic distribution, 3D skeletal posture, and sEMG signals;

[0040] S2: Temporal and spatial alignment of multimodal data to extract muscle thermal gradients and joint trajectory deviation δ;

[0041] S3: When Or when δ>10%, a real-time voice alarm is generated and the training parameters are dynamically adjusted;

[0042] S4: Calculate a personalized metabolic plan based on post-exercise thermal imaging data and push it to the user terminal.

[0043] Furthermore, in S3, dynamically adjusting the training parameters includes:

[0044] Optimize the interval time between high-intensity interval training (HIIT) sessions using the IAGABP algorithm, with an adjustment accuracy of ±5 seconds.

[0045] When the joint trajectory deviates continuously beyond the limit, the output power of the motion device is forced to be reduced to a safe threshold.

[0046] The beneficial effects of the present invention are:

[0047] (1) Through multimodal sensing fusion technology, the system can achieve millisecond-level synchronous detection of abnormal thermodynamic characteristics of muscles and deviations of joint motion trajectories, which can trigger immediate intervention in the initial stage of muscle overstrain or posture instability, significantly reducing the probability of sports injuries.

[0048] (2) Based on three-dimensional skeletal posture reconstruction and bioelectric signal analysis, the system's assessment accuracy of movement standardization reaches competitive level standards, and can accurately identify force timing deviations and joint stability defects, providing a scientific basis for dynamic adjustment of training plans and effectively improving the efficiency of sports skill acquisition.

[0049] (3) By using infrared thermal imaging energy dissipation maps and fusion analysis of physiological parameters, the system can achieve personalized modeling of calorie consumption after exercise, and generate protein supplementation and recovery time recommendations based on this, completely solving the problem of inaccurate nutrition management caused by traditional methods relying on empirical estimation.

[0050] (4) Build an intelligent closed-loop network of school training, home exercise and medical examination data to form a dynamic profile of students' physical health throughout the entire scenario, supporting early prediction and proactive intervention of long-term health risks such as cardiopulmonary endurance deterioration and musculoskeletal strain.

[0051] (5) The system automatically performs movement monitoring, injury warning, and plan optimization tasks, significantly reducing the need for manual supervision, allowing teachers to focus their resources on sports skills guidance and healthy behavior cultivation, and promoting the transformation of physical education to a high-quality, intensive model.

[0052] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0054] Figure 1 This is a system block diagram of the present invention;

[0055] Figure 2 This is the common human body node extraction and posture estimation diagram based on visible light CCD of the present invention;

[0056] Figure 3 This is a human body heat calculation diagram of the infrared image of the present invention. DETAILED DESCRIPTION

[0057] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0058] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0059] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0060] See also Figures 1 to 3 The intelligent sports AI assistant system constructed by the present invention relies on the fusion of multimodal sensing and AI algorithms to realize the full-process intelligent management of students' sports ability and health behavior. The system hardware integrates the FLIR Lepton 3.5 infrared thermal imaging module (resolution 160×120, frame rate 9Hz) and the deep vision skeletal motion recognition unit (including RGB-D camera and TOF sensor, skeletal point positioning accuracy ±2mm), and constructs a temperature-posture synchronization analysis model by real-time collection of human body thermodynamic distribution (temperature measurement range -40℃~550℃) and three-dimensional skeletal trajectory (22 key points). The data is multimodally fused by the Deep Seek-Pro large model (based on Transformer architecture, with 120 million parameters). When the temperature difference gradient of the target muscle group is detected to be greater than 2℃ / cm 2 When the trapezius muscle is overheated (e.g., local overheating) or the joint trajectory deviates from the standard template (Euclidean distance error > 15%), personalized voice instructions are generated immediately (e.g., "The right knee is bent inward beyond the limit, it is recommended to adjust the squat range by 10cm") and transmitted to the sports terminal via Bluetooth 5.0.

[0061] The system synchronously calculates the thermal imaging energy dissipation map after exercise, combines the metabolic equivalent (METs) model to estimate calorie consumption (error <3%), and intelligently pushes protein supplement plans (such as 25g of whey protein for every 500kcal consumed). In terms of sports ability training, the training plan is dynamically optimized based on reinforcement learning (such as dynamically adjusting the interval time of the HIIT group by ±5s), and outputs the action standardization score (0-100 points, including dimensions such as joint stability and force timing). In terms of safety protection, the dual-threshold warning mechanism of abnormal body temperature (local temperature difference >3°C) and posture deviation is used to trigger an audible and visual alarm and force a reduction in exercise load (such as the treadmill slope automatically returns to zero).

[0062] The system integrates training tests, physical examination data and home exercise records to build a digital archive of students' physical health. It uses the temporal convolutional network (TCN) to predict health risks (such as the probability of meniscus wear), achieves proactive injury intervention and cross-scenario safety closed-loop management, and provides a scientific and personalized integrated solution for physical education.

[0063] Example 1: Track and Field Training Injury Real-Time Blocking System

[0064] Workflow:

[0065] 1.1 Synchronous cross-modal data acquisition

[0066] During a sprinter's starting training, the infrared thermal imaging module scans the quadriceps area and generates a temperature distribution matrix T(x,y);

[0067] The RGB-D camera group reconstructs the three-dimensional motion trajectory of the knee joint and outputs the patellar center coordinate P in real time. actual ;

[0068] The TOF sensor and infrared module hardware are triggered synchronously, and the time and space alignment error is controlled within the millisecond level.

[0069] 1.2 Dual Threshold Risk Decision

[0070] Thermal gradient calculation:

[0071]

[0072] Among them, T i,j is the temperature value of the pixel in the i-th row and j-th column of the infrared thermal imaging image, in degrees Celsius; Δx is the physical distance between adjacent pixels in the horizontal direction, in centimeters; Δy is the physical distance between adjacent pixels in the vertical direction, in centimeters; is the maximum temperature gradient in the target area, in degrees Celsius per square centimeter.

[0073] Proximal vastus lateralis muscle

[0074] Joint deviation calculation:

[0075]

[0076] Among them, P actual is the three-dimensional coordinate vector of the joint key points collected in real time, in millimeters; P standard is the three-dimensional coordinate vector of the corresponding joint point in the standard action template, in millimeters; ||·||2 is the Euclidean norm of the vector, that is, the modulus; δ is the deviation of the joint trajectory, in percentage.

[0077] The measured starting and stretching phase δ = 12%;

[0078] The DeepSeek-Pro model integrates two parameters, triggering a Level 1 alarm.

[0079] 1.3 Real-time intervention execution

[0080] AR glasses projection correction guide: "Reduce the ground push angle by 5°";

[0081] Smart running shoes automatically adjust the stiffness of the heel cushioning;

[0082] The system records this event in the sports injury risk map.

[0083] Example 2: Fitness Metabolic Actuarial and Nutrition Closed-Loop System

[0084] 2.1 Multimodal energy consumption monitoring

[0085] After the user completes resistance training, infrared thermal imaging scans the latissimus dorsi area:

[0086] Grid unit dA = 1 cm 2 ;

[0087] Record the temperature difference ΔT(x,y) for each unit.

[0088] 2.2 Metabolic optimization calculation

[0089] Calorie expenditure modeling:

[0090]

[0091] Where, ΔT i is the temperature difference between the skin surface and the basal body temperature on the i-th grid unit, in degrees Celsius; dA i is the area of ​​the i-th grid unit, in square centimeters; n is the total number of divided grid units; k is the heat dissipation energy conversion coefficient, in kcal / (℃·cm 2 ), determined by calibration experiments; MET is metabolic equivalent of task, which indicates the multiple of exercise intensity relative to resting metabolism; b is metabolic equivalent calibration coefficient, unit is kilocalorie (kcal); E cal is the total calorie consumption value in kilocalories (kcal). k = 0.38, b = 1.2 is determined by calibration;

[0092] Protein recommendations:

[0093] M protein =0.05·E cal +β

[0094] 0.05 is the protein supplement coefficient, in grams per kilocalorie (g / kcal); β is the personalized correction term, in grams (g), which is dynamically adjusted based on the user's body composition data; M protein The recommended protein intake is expressed in grams (g).

[0095] According to the user's body fat percentage, β is set to -2g, and the recommended value is output as 28g.

[0096] 2.3 Dynamic Training Adjustment

[0097] IAGABP algorithm optimizes the next day plan:

[0098] Chromosome encoding: load weight, group rest time;

[0099] Fitness function: in, The maximum temperature gradient value of the monitoring area in a single training session, in degrees Celsius per square centimeter (℃ / cm 2 ); f is the chromosome fitness score (dimensionless), with a value range of (0,1].

[0100] Output squat load reduction program.

[0101] 2.4 Cross-terminal execution

[0102] The home smart refrigerator receives the protein supplement instruction and lights up to indicate the location of the protein drink;

[0103] Gym equipment automatically adjusts weight stacks to the new program.

[0104] Example 3: Adolescent Spine Health Full-Cycle Management System

[0105] Scenario: Scoliosis screening and intervention project in key middle schools

[0106] 3.1 Learning Scenario Monitoring

[0107] Students study at their desks for two hours:

[0108] Infrared thermal imaging continuously monitors the erector spinae muscle temperature field T(x,y,t);

[0109] Calculate local metabolic equivalents:

[0110]

[0111] in, is the rate of change of the target muscle area temperature over time, in degrees Celsius per minute (°C / min); α is the thermal metabolism conversion coefficient, in kilocalories per degree Celsius (kcal / °C); dA is the microelement area of ​​the muscle area, in square centimeters (cm 2 );∫ Ais the integral of the target muscle area A; BMR is the user's basal metabolic rate, in kcal / day; MET local It is the metabolic equivalent of local muscle activity.

[0112] 3.2 Health Risk Prediction

[0113] Temporal convolutional network input features:

[0114]

[0115] Output Scoliosis Progression Probability:

[0116] P risk =TCN(X t-6 :X t )

[0117] Among them, X t is the feature vector at time t, including: is the temperature gradient of the erector spinae muscles on the left and right sides of the spine, expressed in degrees Celsius per centimeter (℃ / cm); δ Cobb is the Cobb angle deviation value of the spine three-dimensional reconstruction, in degrees (°); The cumulative calorie consumption for the day, in kilocalories (kcal);

[0118] TCN(·) is a temporal convolutional neural network model; P risk The probability of scoliosis progression in the next six months, ranging from [0,1].

[0119] When P risk >0.7 triggers an early warning.

[0120] 3.3 Cross-scenario intervention closed loop

[0121] School side: Automatically adjust the support surface of desks and chairs;

[0122] Home: VR system pushes Schroth gymnastics courses;

[0123] Medical side: Generate 3D printing orthotic brace parameters.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture, characterized by: include: A multimodal motion data perception module configured to simultaneously collect human infrared thermal imaging data, three-dimensional skeletal posture data, and surface electromyography (sEMG) signals; An AI-driven training optimization engine, in communication with the multimodal motion data perception module, for generating real-time motion instructions and dynamic training plans based on multimodal data fusion analysis; The full-scenario intelligent management platform is connected to the AI-driven training optimization engine to perform injury warning, metabolic optimization and cross-scenario closed-loop management of health data.

2. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 1 is characterized by: The multimodal motion data perception module includes: Infrared thermal imaging unit, using an infrared sensor with a resolution of no less than 160×120 and a temperature measurement range of -40°C to +550°C; A 3D skeleton reconstruction unit, including an RGB-D camera and a Time of Flight (TOF) sensor, is configured to extract at least 18 key points of the human body and achieve 3D reconstruction of the skeleton posture with a positioning accuracy of ≤±2mm; The bioelectric signal acquisition unit is configured to synchronously acquire surface electromyography (sEMG) signals.

3. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 1 is characterized by: The AI-driven training optimization engine performs the following operations: (a) Establish a dual-threshold judgment model for muscle thermal gradient and joint trajectory deviation: When the temperature gradient in the target area is detected Or when the joint trajectory deviation δ>10%, a real-time adjustment instruction is generated; in, T is the muscle surface temperature, s is the spatial coordinate; P is the joint coordinate; (b) Dynamically generate a phased training plan based on the Improved Adaptive Genetic Algorithm Backpropagation (IAGABP).

4. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 1 is characterized by: The full-scenario intelligent management platform includes: The damage warning submodule is configured to trigger an audible and visual alarm or forced load reduction operation within 100ms; Metabolic optimization submodule calculates calorie consumption value E based on post-exercise thermal imaging data cal : E cal =k·ζ A ΔT(x,y)dA+b·MET Wherein, ΔT is the temperature difference distribution on the body surface, A is the effective heat dissipation area, MET is the metabolic equivalent, k and b are calibration coefficients; The data closed-loop sub-module integrates school training, home exercise and physical examination data to build a dynamic physical health file.

5. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 4 is characterized by: The metabolic optimization submodule is based on the calorie consumption value E cal Recommended protein intake M protein : M protein =α·E cal +b Here, α=0.05g / kcal, and β is a correction term based on the user's physique.

6. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 2 is characterized by: The infrared thermal imaging unit and the three-dimensional bone reconstruction unit adopt a hardware synchronization trigger mechanism, and the time-space alignment error is ≤1ms.

7. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 3 is characterized by: The dual-threshold judgment model is implemented through the DeepSeek-Pro motion optimization model, which adopts the Transformer architecture and has a parameter volume of ≥120 million.

8. The real-time early warning system for sports injuries and health optimization based on the fusion of multimodal infrared thermal imaging and three-dimensional skeletal posture according to claim 1 is characterized by: The real-time motion instructions are output through augmented reality AR glasses or bone conduction headphones, and the instruction content includes a load adjustment range of ±15%.

9. A sports injury early warning and health optimization method based on the system of any one of claims 1 to 8, characterized in that: Including steps: S1: Synchronously collects the user's thermodynamic distribution, 3D skeletal posture, and sEMG signals; S2: Temporal and spatial alignment of multimodal data to extract muscle thermal gradients and joint trajectory deviation δ; S3: When Or when δ>10%, a real-time voice alarm is generated and the training parameters are dynamically adjusted; S4: Calculate a personalized metabolic plan based on post-exercise thermal imaging data and push it to the user terminal.

10. The sports injury early warning and health optimization method according to claim 9, characterized in that: In S3, dynamically adjusting the training parameters includes: Optimize the interval time between high-intensity interval training (HIIT) sessions using the IAGABP algorithm, with an adjustment accuracy of ±5 seconds. When the joint trajectory deviates continuously beyond the limit, the output power of the motion device is forced to be reduced to a safe threshold.

Images