Method and system for sensing motion posture of moving body

Abstract

The present invention relates to the field of motion posture perception technology, and in particular to a motion posture perception method and system for a moving body. The method comprises the following steps: real-time shooting of motion images corresponding to the motion process of the moving body to obtain real-time motion change image data of the moving body; real-time monitoring of motion state parameters corresponding to the motion process of the moving body to obtain real-time acceleration data and real-time angular velocity data of the moving body; performing time sequence alignment and synchronization processing, and simultaneously performing motion coordinate positioning and long-term and short-term complementary filtering fusion analysis of the moving body to obtain a fusion feature sequence of motion posture change of the moving body; based on the fusion feature sequence of motion posture change of the moving body, performing motion posture trajectory recognition and analysis of the motion process, and simultaneously performing posture change pattern recognition and analysis and motion posture perception classification to obtain the corresponding motion posture category information of the moving body in the current motion process. The present invention can realize high-precision perception of the motion posture of the moving body.

Description

Technical Field

[0001] The present invention relates to the technical field of motion posture perception, and in particular to a motion posture perception method and system for a moving body. Background Art

[0002] The motion posture perception method of a moving body aims to monitor and analyze the posture of an athlete in real time to improve the effect of exercise, prevent sports injuries, and provide scientific guidance for sports training. By combining visual sensors and inertial sensors, a comprehensive data acquisition and processing framework is established to obtain the posture information of the athlete in real time. In the process of data fusion, advanced filtering algorithms (such as Kalman filtering) are used to reduce sensor noise and errors, thereby improving the accuracy and stability of posture recognition. The training model is also used to analyze the athlete's historical data, identify individual characteristics and movement patterns, realize personalized posture evaluation and feedback, and provide targeted improvement suggestions to help athletes better adjust their postures in subsequent training and optimize the exercise effect. However, the existing motion posture perception methods mainly rely on cameras and visual sensors. In complex motion scenes or environments with large changes in lighting, these methods will cause the visual sensors to be blocked or reflected to produce corresponding measurement noise and drift, which often easily leads to cumulative errors, resulting in reduced posture recognition accuracy. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a method and system for sensing the motion posture of a moving object to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above object, a method for sensing the motion posture of a moving object comprises the following steps:

[0005] Step S1: Real-time motion image capture of the motion process corresponding to the moving body is performed through a high-resolution camera to obtain real-time motion change image data of the moving body; real-time monitoring of motion state parameters of the motion process corresponding to the moving body is performed through a built-in accelerometer and gyroscope to obtain real-time acceleration data of the moving body and real-time angular velocity data of the moving body; time-series alignment and synchronization processing is performed on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body and the real-time angular velocity data of the moving body to obtain motion image data of the moving body, acceleration data of the moving body and angular velocity data of the moving body in the same time-series dimension;

[0006] Step S2: positioning the motion coordinates of the moving body on the moving image data to obtain a set of the moving body's motion space coordinate points; performing a long-term and short-term complementary filtering fusion analysis on the moving body's acceleration data and the moving body's angular velocity data based on the moving body's motion space coordinate point set to obtain a fusion feature sequence of the moving body's motion posture change;

[0007] Step S3: Based on the fusion feature sequence of the motion posture change of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition and analysis to generate the motion posture change trajectory of the moving body;

[0008] Step S4: performing posture change pattern recognition analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; performing motion posture perception classification on the corresponding motion posture change trajectory of the moving body based on the motion posture change pattern of the moving body to obtain the corresponding motion posture category information of the moving body in the current motion process.

[0009] Further, step S1 includes the following steps:

[0010] Step S11: Using a high-resolution camera to take real-time motion image capture of the motion process of the moving object, and obtaining real-time motion change image data of the moving object;

[0011] Step S12: Real-time monitoring of the motion acceleration of the motion object corresponding to the motion process is performed by a built-in accelerometer to obtain real-time acceleration data of the motion object;

[0012] Step S13: monitoring the angular velocity of the motion process of the moving body in real time through the built-in gyroscope to obtain the real-time angular velocity data of the moving body;

[0013] Step S14: performing data smoothing preprocessing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body, and the real-time angular velocity data of the moving body to obtain the smoothed data of the motion change image of the moving body, the smoothed data of the acceleration of the moving body, and the smoothed data of the angular velocity of the moving body;

[0014] Step S15: performing time-series alignment and synchronization processing on the moving body motion change image smoothing data, the moving body acceleration smoothing data and the moving body angular velocity smoothing data to obtain the moving body motion image data, the moving body acceleration data and the moving body angular velocity data in the same time-series dimension.

[0015] Further, step S2 includes the following steps:

[0016] Step S21: extracting motion image corner points from each continuous frame of motion image data of the moving object to obtain a set of motion image feature corner points of the moving object;

[0017] Step S22: Establishing a three-dimensional space coordinate system for the moving object motion image data to generate a three-dimensional space coordinate system for the moving object motion image;

[0018] Step S23: performing image space position geometric transformation calculation on each continuous frame of the moving image data of the moving object based on the moving image three-dimensional space coordinate system to obtain the moving image space position geometric registration matrix;

[0019] Step S24: positioning the moving body motion coordinates of each corresponding moving image feature corner point in the moving body motion image feature corner point set based on the moving body motion image spatial position geometric registration matrix to obtain the moving body motion spatial coordinate point set;

[0020] Step S25: Based on the moving body's moving space coordinate point set, long-term and short-term complementary filtering fusion analysis is performed on the moving body's acceleration data and the moving body's angular velocity data to obtain a moving body's moving posture change fusion feature sequence.

[0021] Further, step S23 includes the following steps:

[0022] Step S231: generating image coordinate points for each continuous frame of motion image in the motion image data of the moving object, and obtaining a set of motion image coordinate points of the moving object corresponding to each frame of motion image;

[0023] Step S232: performing adjacent frame space geometric transformation estimation on the moving body motion image coordinate point set corresponding to each frame of motion image based on the moving body motion image three-dimensional space coordinate system, so as to obtain image space geometric transformation parameters corresponding to each adjacent frame of motion image, wherein the image space geometric transformation parameters include image space position transformation parameters, image space rotation transformation parameters and image space scaling transformation parameters;

[0024] Step S233: performing geometric registration matrix transformation calculation according to the image space geometric transformation parameters corresponding to each adjacent frame of motion images to obtain a geometric registration matrix of the motion image space position of the moving object.

[0025] Further, step S25 includes the following steps:

[0026] Step S251: performing point matching and division on the acceleration data and angular velocity data of the moving body based on each moving space coordinate point in the moving body moving space coordinate point set, and obtaining the moving body acceleration and moving body angular velocity corresponding to each moving space coordinate point;

[0027] Step S252: monitoring the dynamic response duration of the moving body for each moving space coordinate point in the moving body moving space coordinate point set, and obtaining the corresponding motion response duration at each moving space coordinate point;

[0028] Step S253: According to the preset response time threshold, the motion response time corresponding to each motion space coordinate point is traversed point by point to determine; when the motion response time is greater than or equal to the preset response time threshold, the corresponding motion space coordinate point is marked as a long-term response coordinate point; when the motion response time is less than the preset response time threshold, the corresponding motion space coordinate point is marked as a short-term response coordinate point;

[0029] Step S254: merging the long-term motion state parameters of the moving body acceleration and the moving body angular velocity corresponding to the long-term response coordinate point marked as the moving body to obtain the moving body long-term response motion state parameter subset; merging the short-term motion state parameters of the moving body acceleration and the moving body angular velocity corresponding to the short-term response coordinate point marked as the moving body to obtain the moving body short-term response motion state parameter subset; performing motion posture change feature analysis on the moving body long-term response motion state parameter subset and the moving body short-term response motion state parameter subset respectively to obtain the moving body long-term motion posture change feature subset and the moving body short-term motion posture change feature subset;

[0030] Step S255: performing long-term and short-term complementary filtering fusion analysis on the long-term motion posture change feature subset of the moving body and the short-term motion posture change feature subset of the moving body to obtain a fusion feature sequence of the motion posture change of the moving body.

[0031] Further, step S255 includes the following steps:

[0032] Separating the dynamic and static features of the long-term motion posture change feature subset and the short-term motion posture change feature subset of the moving body respectively, and obtaining the dynamic feature component of the long-term motion posture of the moving body, the static feature component of the long-term motion posture of the moving body, the dynamic feature component of the short-term motion posture of the moving body, and the static feature component of the short-term motion posture of the moving body;

[0033] The characteristic component cumulative error calculation formula is used to calculate the long-term and short-term cumulative errors of the dynamic characteristic component of the long-term motion posture of the moving body, the static characteristic component of the long-term motion posture of the moving body, the dynamic characteristic component of the short-term motion posture of the moving body, and the static characteristic component of the short-term motion posture of the moving body, so as to obtain the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body;

[0034] Based on the accumulated error of the long-term motion posture of the moving body and the accumulated error of the short-term motion posture of the moving body, the long-term motion posture change feature subset and the short-term motion posture change feature subset of the moving body are subjected to long-term and short-term complementary filtering fusion analysis to obtain the fusion feature sequence of the moving body's motion posture change.

[0035] Furthermore, the characteristic component cumulative error calculation formula is specifically:

[0036] ;

[0037] ;

[0038] In the formula, is the long-term cumulative error of the motion posture of the moving body, is the cumulative error of the short-term motion posture of the moving body, is the total time of long-term motion posture observation, is the total time of short-term motion posture observation, is the integral time variable parameter, is the total number of long-term motion posture features in the long-term motion posture change feature subset of the moving body, is the item index parameter of the long-term motion posture feature, For in time The corresponding dynamic characteristic reference value of the moving body's motion posture, For in time The static characteristic reference value of the motion posture of the moving body corresponding to the position, For the Long-term motion gesture features in time The corresponding long-term motion posture dynamic characteristic components of the moving body, is the error weight coefficient of the long-term dynamic characteristic component, For the Long-term motion gesture features in time The corresponding static characteristic components of the long-term motion posture of the moving body, is the error weight coefficient of the long-term static characteristic component, is the total number of short-term motion posture features in the short-term motion posture change feature subset of the moving body, is the item index parameter of short-term motion posture feature, For the Short-term motion gesture features in time The corresponding short-term motion posture dynamic characteristic components of the moving body, is the error weight coefficient of the short-term dynamic characteristic component, For the Short-term motion gesture features in time The corresponding static characteristic component of the short-term motion posture of the moving body, is the error weight coefficient of the short-term static characteristic component, is the correction coefficient of the cumulative error of the long-term motion posture of the moving body, It is the correction coefficient of the cumulative error of the moving body’s short-term motion posture.

[0039] Further, the long-term and short-term complementary filtering fusion analysis of the long-term motion posture change feature subset and the short-term motion posture change feature subset of the moving body based on the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body includes the following steps:

[0040] Compare and judge the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body. If the long-term motion posture cumulative error of the moving body is greater than or equal to the short-term motion posture cumulative error of the moving body, set the short-term motion posture feature sliding window according to the difference between the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body, and perform short-term collaborative filtering on the short-term motion posture change feature subset of the moving body based on the short-term motion posture feature sliding window to obtain the corresponding short-term motion posture filtering feature component of the moving body in each short-term sliding window;

[0041] The short-term motion posture filtering feature component of the moving body corresponding to each short-term sliding window is short-term replaced and complemented with the long-term motion posture feature component of the moving body corresponding to the long-term motion posture change feature subset in the same short-term sliding window to obtain a complementary sequence of long-term motion posture features of the moving body;

[0042] If the long-term motion posture cumulative error of the moving body is less than the short-term motion posture cumulative error of the moving body, a long-term motion posture feature sliding window is set according to the difference between the short-term motion posture cumulative error of the moving body and the long-term motion posture cumulative error of the moving body, and a long-term collaborative filtering is performed on the long-term motion posture change feature subset of the moving body based on the long-term motion posture feature sliding window to obtain the corresponding long-term motion posture filtering feature component of the moving body in each long-term sliding window;

[0043] The long-term motion posture filtering feature component of the moving body corresponding to each long-term sliding window is subjected to long-term replacement and complementation with the short-term motion posture feature component of the moving body corresponding to the short-term motion posture change feature subset in the same long-term sliding window to obtain a complementary sequence of the short-term motion posture features of the moving body;

[0044] The dynamic time warping method is used to perform long-term and short-term dynamic planning fusion analysis on the long-term motion posture feature complementary sequence and the short-term motion posture feature complementary sequence of the moving body, and the fusion feature sequence of the moving body's motion posture change is obtained.

[0045] Further, step S3 includes the following steps:

[0046] Step S31: performing motion process parameter description analysis on the motion posture change fusion feature sequence of the moving body to obtain an actual parameter description set of the motion process of the moving body, wherein the actual parameter description set of the motion process of the moving body includes actual change parameters of the acceleration of the moving body and actual change parameters of the angular velocity of the moving body;

[0047] Step S32: Based on the actual parameterized description set of the motion process of the moving body, the corresponding motion process is subjected to motion key node identification processing to obtain a motion key node set of the moving body motion process, wherein the motion key node set of the moving body motion process includes a starting point, a turning point and an end point;

[0048] Step S33: Based on the actual parameterized description set of the motion process of the moving body and the key node set of the motion process of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition analysis to generate the motion posture change trajectory of the moving body.

[0049] Furthermore, the present invention also provides a moving body motion posture perception system, which is used to execute the moving body motion posture perception method as described above, and the moving body motion posture perception system comprises:

[0050] The motion state parameter analysis module of the moving body is used to take real-time motion images of the motion process corresponding to the moving body through a high-resolution camera to obtain real-time motion change image data of the moving body; to monitor the motion state parameters of the motion process corresponding to the moving body through a built-in accelerometer and gyroscope in real time to obtain real-time acceleration data of the moving body and real-time angular velocity data of the moving body; to perform time-series alignment and synchronization processing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body and the real-time angular velocity data of the moving body, so as to obtain the motion image data of the moving body, the acceleration data of the moving body and the angular velocity data of the moving body in the same time-series dimension;

[0051] The moving body motion feature filtering and fusion module is used to locate the moving body motion coordinates of the moving body motion image data to obtain the moving body motion space coordinate point set; based on the moving body motion space coordinate point set, the moving body acceleration data and the moving body angular velocity data are subjected to long-term and short-term complementary filtering fusion analysis to obtain the moving body motion posture change fusion feature sequence;

[0052] A motion body motion posture trajectory recognition module is used to perform motion posture trajectory recognition and analysis on the corresponding motion process based on the motion body motion posture change fusion feature sequence to generate the motion body motion posture change trajectory;

[0053] The motion state perception and classification module of the moving body is used to perform posture change pattern recognition and analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; based on the motion posture change pattern of the moving body, the corresponding motion posture change trajectory of the moving body is subjected to motion posture perception and classification to obtain the corresponding motion posture category information of the moving body in the current motion process.

[0054] Beneficial effects of the present invention:

[0055] 1. The motion posture perception method of the moving body proposed in the present invention, compared with the prior art, has the beneficial effect that by using a high-resolution camera to perform real-time motion image shooting of the motion process corresponding to the moving body, high-quality motion image data can be obtained. This real-time monitoring provides rich visual information for the motion analysis of the moving body, and can clearly record every subtle change of the moving body, which not only helps to analyze the motion trajectory, posture and behavior, but also provides basic data for subsequent action recognition and classification. By using a built-in accelerometer to perform real-time monitoring of the motion acceleration of the motion process corresponding to the moving body, the acceleration change of the moving body can be monitored in real time. This process can provide important data for analyzing the intensity, acceleration and deceleration of the movement. By monitoring the acceleration, researchers can understand the dynamic changes experienced by the moving body in different motion stages, and provide data support for the subsequent processing process. Also, by using a built-in gyroscope to perform real-time monitoring of the motion angular velocity of the motion process corresponding to the moving body, it can be used to monitor the angular velocity change of the moving body in real time, and can provide the rotation and rotation information of the moving body in three-dimensional space, thereby providing basic data guarantee for subsequent motion posture perception analysis. At the same time, by performing time-series alignment and synchronization processing on the real-time motion change image data, real-time acceleration data and real-time angular velocity data of the moving body, this process can integrate data from different sources into the same time-series dimension, providing a unified basis for comprehensive motion analysis. Time-series alignment can help analyze the relationship between the various parameters of the moving body and reveal the inherent laws in the motion process. For example, the changing relationship between the speed, acceleration and angular velocity of the moving body can be analyzed more accurately, helping the moving body to optimize its motion techniques and tactics, making the subsequent motion posture perception analysis process more comprehensive and in-depth. Secondly, by locating the motion coordinates of the moving body's motion image data, each feature point of the moving body in the three-dimensional space can be accurately located. The core of this process is to combine image data with three-dimensional space, so as to realize the positioning of the feature corner points of the moving body in the real three-dimensional space. The realization of this step not only improves the accuracy of motion analysis, but also makes the subsequent analysis of dynamic features such as motion trajectory, speed and acceleration possible. By performing long-term and short-term complementary filtering fusion analysis on the acceleration data and angular velocity data of the moving body based on the spatial coordinate point set of the moving body, the accuracy and reliability of the motion posture changes can be effectively improved. The long-term and short-term complementary filtering technology combines the short-term motion posture feature information and the long-term motion posture feature information, and can capture the rapid changes of the moving body in real time, and can also smooth the slow-changing trends. This analysis method has significant advantages in processing dynamic systems, and can effectively suppress the influence of noise, reduce the cumulative error between the long-term and short-term motion posture feature information, and improve the stability and accuracy of subsequent posture estimation, thereby helping researchers and engineers to deeply understand the dynamic behavior of the moving body.Then, by performing motion posture trajectory recognition and analysis on the corresponding motion process based on the fusion feature sequence of the motion posture change of the motion body, this process can form a panoramic understanding of the posture change of the motion body during the motion process, which helps to grasp the motion characteristics and behavior patterns of the motion body at different stages. By analyzing the motion trajectory, the motion laws and skills of the motion body can be identified, thereby providing a scientific basis for subsequent motion posture perception. Finally, by performing posture change pattern recognition and analysis on the motion posture change trajectory of the motion body, the posture change pattern of the motion body in the current motion process can be obtained, which can help researchers deeply understand the performance of the motion body at different stages and the laws of its motion posture change. This process can reveal the different postures adopted by athletes in sports and how these postures affect sports performance by analyzing the characteristics of the motion trajectory. In addition, by performing motion posture perception classification on the corresponding motion posture change trajectory of the moving body based on the motion posture change pattern of the moving body, the motion posture category information of the moving body in the current motion process can be obtained, and the results of motion recognition can be converted into actionable information to provide support for the behavioral analysis of the moving body. This step can help researchers accurately understand the current motion posture state and possible motion effects of the moving body, including standing, walking, jumping, etc., through the classification of motion postures, and then make timely adjustments and improvements. This classification process can also realize personalized training plans and customize corresponding exercise plans according to the needs and abilities of different athletes, thereby improving the accuracy of motion posture recognition.

[0056] 2. The motion posture perception system of a moving body proposed in the present invention is generally composed of a motion state parameter analysis module of a moving body, a motion feature filtering and fusion module of a moving body, a motion posture trajectory recognition module of a moving body and a motion state perception classification module. It can realize any motion posture perception method of a moving body described in the present invention, and is used to combine the operations between computer programs running on various modules to realize the motion posture perception method of a moving body. The internal structures of the system cooperate with each other, which can greatly reduce duplication of work and manpower investment, and can quickly and effectively provide a more accurate and efficient motion posture perception process of a moving body, thereby simplifying the operation flow of the motion posture perception system of a moving body. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0058] Figure 1 It is a schematic diagram of the steps of the method for sensing the motion posture of a moving body according to the present invention;

[0059] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0060] Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0061] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0062] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0063] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0064] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for sensing the motion posture of a moving body, the method comprising the following steps:

[0065] Step S1: Real-time motion image capture of the motion process corresponding to the moving body is performed through a high-resolution camera to obtain real-time motion change image data of the moving body; real-time monitoring of motion state parameters of the motion process corresponding to the moving body is performed through a built-in accelerometer and gyroscope to obtain real-time acceleration data of the moving body and real-time angular velocity data of the moving body; time-series alignment and synchronization processing is performed on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body and the real-time angular velocity data of the moving body to obtain motion image data of the moving body, acceleration data of the moving body and angular velocity data of the moving body in the same time-series dimension;

[0066] Step S2: positioning the motion coordinates of the moving body on the moving image data to obtain a set of the moving body's motion space coordinate points; performing a long-term and short-term complementary filtering fusion analysis on the moving body's acceleration data and the moving body's angular velocity data based on the moving body's motion space coordinate point set to obtain a fusion feature sequence of the moving body's motion posture change;

[0067] Step S3: Based on the fusion feature sequence of the motion posture change of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition and analysis to generate the motion posture change trajectory of the moving body;

[0068] Step S4: performing posture change pattern recognition analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; performing motion posture perception classification on the corresponding motion posture change trajectory of the moving body based on the motion posture change pattern of the moving body to obtain the corresponding motion posture category information of the moving body in the current motion process.

[0069] In the embodiment of the present invention, please refer to Figure 1 FIG. 1 is a schematic diagram of the steps of the motion posture sensing method of a moving body according to the present invention. In this example, the motion posture sensing method of a moving body comprises the following steps:

[0070] Step S1: Real-time motion image capture of the motion process corresponding to the moving body is performed through a high-resolution camera to obtain real-time motion change image data of the moving body; real-time monitoring of motion state parameters of the motion process corresponding to the moving body is performed through a built-in accelerometer and gyroscope to obtain real-time acceleration data of the moving body and real-time angular velocity data of the moving body; time-series alignment and synchronization processing is performed on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body and the real-time angular velocity data of the moving body to obtain motion image data of the moving body, acceleration data of the moving body and angular velocity data of the moving body in the same time-series dimension;

[0071] In an embodiment of the present invention, the motion process of the moving body is photographed in real time by using a high-resolution camera. The selected high-resolution camera needs to have a resolution of at least 1080p and can automatically adjust the exposure under different lighting conditions to ensure that a clear motion image is captured. The camera is installed at an appropriate position of the moving body to ensure that the camera field of view can cover the entire motion trajectory of the moving body, thereby obtaining the real-time motion change image data of the moving body. The motion process corresponding to the moving body is monitored in real time by using a built-in accelerometer. The accelerometer used needs to have a three-axis measurement function and a high sampling frequency (e.g., 100 Hz) to capture the acceleration changes of the moving body on the X, Y, and Z axes in real time, thereby obtaining the real-time acceleration data of the moving body. At the same time, the angular velocity of the motion process corresponding to the moving body is monitored in real time by using a built-in gyroscope. The selected gyroscope needs to be able to measure the angular velocity in three-dimensional space and have high sensitivity and low drift characteristics. After the moving body is started, the gyroscope begins to measure its angular velocity changes in the three axes of X, Y, and Z, thereby obtaining the real-time angular velocity data of the moving body. Then, by comparing the timestamps in each data source of the real-time motion change image data of the moving body, the real-time acceleration data of the moving body, and the real-time angular velocity data of the moving body previously obtained through real-time monitoring, it is ensured that the motion change image data, acceleration data, and angular velocity data of the moving body are compared and analyzed in the same time dimension, and the time difference between the data is identified therefrom, and the acceleration data and angular velocity data are time-resampled by applying an interpolation algorithm (such as linear interpolation or spline interpolation) to align their timestamps with the motion change image data. In this way, it is ensured that all data are consistent in the same time dimension, and finally the motion image data, acceleration data, and angular velocity data of the moving body in the same time series dimension are obtained.

[0072] Step S2: positioning the motion coordinates of the moving body on the moving image data to obtain a set of the moving body's motion space coordinate points; performing a long-term and short-term complementary filtering fusion analysis on the moving body's acceleration data and the moving body's angular velocity data based on the moving body's motion space coordinate point set to obtain a fusion feature sequence of the moving body's motion posture change;

[0073] In the embodiment of the present invention, the optical flow method is used in the motion image data of the moving body to extract corner points of the motion image of each continuous frame. The optical flow method identifies motion features by calculating the change in pixel intensity between adjacent frames, and selects pixels with large changes as feature corner points. A three-dimensional spatial coordinate system is established for the motion image data of the moving body. First, the origin position of the reference coordinate system is determined. Usually, the position of the moving body in a stationary state is selected as the origin. Three mutually perpendicular coordinate axes (X, Y, and Z axes) are set to represent the motion of the moving body in different directions, respectively. A three-dimensional reconstruction technology is used, and motion images of multiple perspectives are combined. A camera calibration algorithm is used to obtain Take the camera's internal and external parameters to establish a three-dimensional coordinate system. At the same time, by combining the previously established three-dimensional space coordinate system, perform image space position geometric transformation calculations on each consecutive frame in the moving image data of the moving object, so as to convert the two-dimensional image coordinates of each frame into three-dimensional space coordinates by applying the perspective transformation algorithm. By constructing a transformation matrix, the coordinate data of the moving image is matched with the corresponding points in the three-dimensional space coordinate system to obtain a geometric registration matrix. In the registration process, the least squares optimization algorithm is used to minimize the projection transformation parameters between consecutive frames, including position transformation, rotation transformation and scaling transformation parameters, and the transformation parameters are combined into a 4x4 homogeneous transformation matrix. , the geometric transformation matrix of each frame is used to geometrically register the moving image of the moving object, and each pair of adjacent frame matrices are multiplied using linear algebra tools (such as the NumPy library) to obtain the geometric registration matrix. After obtaining the geometric registration matrix, the moving image feature corner point set is used to locate the moving coordinates of the moving object based on the matrix, so that for each feature corner point, the corresponding coordinates in the three-dimensional space coordinate system are calculated using the geometric transformation relationship. In the specific operation, the two-dimensional image coordinates of the corner point are transformed in combination with the registration matrix and mapped to the three-dimensional space to ensure that the coordinate positioning of the moving object in space is accurate and continuous, thereby obtaining the moving space coordinate point set of the moving object. Then, by combining the motion space coordinate point set of the moving body obtained by the previous registration transformation, the acceleration data and angular velocity data of the moving body at the corresponding coordinate points are subjected to long-term and short-term complementary filtering fusion analysis, so as to smooth the acceleration data by using a low-pass filter to eliminate the influence of high-frequency noise, and apply a high-pass filter to the angular velocity data to extract its rapidly changing features. Then, the Kalman filter framework is adopted to combine the acceleration and angular velocity data, and through state prediction and updating steps, a characteristic sequence of the motion posture change of the moving body is fused to form. During the analysis process, the time series changes of the fused data are monitored in real time to obtain the posture information of the moving body at different time points, so as to achieve high precision and real-time perception of the moving body posture, and finally obtain a fused characteristic sequence of the motion posture change of the moving body.

[0074] Step S3: Based on the fusion feature sequence of the motion posture change of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition and analysis to generate the motion posture change trajectory of the moving body;

[0075] In an embodiment of the present invention, a parameterized description analysis of the motion process is performed on the motion posture change fusion feature sequence of the moving body obtained after filtering and fusion, so as to calculate the actual change parameters of acceleration and angular velocity using the numerical integration method for the motion data (such as acceleration, angular velocity, position, etc.), and the corresponding motion process is processed by identifying the key nodes of the motion by combining the parameterized description set obtained by the previous analysis, so as to determine the motion state of the moving body corresponding to the motion process by analyzing the acceleration and angular velocity change data of the moving body. In this process, the data is classified by the set threshold value to identify the starting point, turning point and end point. The starting point is determined by the sudden change of acceleration and angular velocity, the turning point is determined according to the severity of the change of acceleration and angular velocity, and the end point is determined by the acceleration. And the angular velocity suddenly changes to zero. Then, the corresponding motion process is identified by combining the parameterized description set obtained by the previous analysis and the key node set of the motion process of the moving body. The state information (such as position, velocity, acceleration, angular velocity, etc.) of the moving body at the key nodes is extracted by using the parameterized description data and key nodes obtained in the previous step, and the interpolation algorithm (such as spline interpolation or linear interpolation) is used to smooth these state information to generate a continuous motion trajectory, analyze the parameter changes of the moving body in different time periods, and accurately describe the posture trajectory of the moving body through data fitting methods (such as polynomial fitting or Bezier curve fitting), which will comprehensively reflect the changes of the moving body in the entire motion process, and finally identify and fit the motion posture change trajectory of the moving body.

[0076] Step S4: performing posture change pattern recognition analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; performing motion posture perception classification on the corresponding motion posture change trajectory of the moving body based on the motion posture change pattern of the moving body to obtain the corresponding motion posture category information of the moving body in the current motion process.

[0077] In an embodiment of the present invention, the posture change pattern is identified and analyzed by analyzing the previously identified and fitted motion posture change trajectory of the moving body, and the motion posture change trajectory of the moving body is analyzed by using a clustering algorithm (such as K-means or hierarchical clustering) to identify the posture change patterns of the moving body at different stages. By extracting feature vectors to describe the key attributes of each posture (such as speed and acceleration changes), a pattern recognition model is established, and a trained classifier is used to identify the characteristic patterns of the moving body during the motion process, and a detailed description of the motion posture change pattern of the moving body is generated to ensure a comprehensive understanding and accurate identification of the posture changes, thereby obtaining the motion posture change pattern of the moving body. At the same time, after obtaining the motion posture change pattern of the moving body, the motion posture perception classification is performed on the corresponding motion posture change trajectories of the moving body, so as to extract features of the identified posture change pattern based on the existing motion posture classification model and construct a feature vector. Then, a deep learning algorithm (such as a convolutional neural network) is used to train and classify the feature vector, and the posture changes of the current motion process are compared with the training set, so as to output the motion posture category information corresponding to the moving body, including standing, walking, jumping, etc., to ensure that the motion state of the moving body can be fed back in real time, and finally the motion posture category information corresponding to the moving body in the current motion process is obtained.

[0078] Further, step S1 includes the following steps:

[0079] Step S11: Using a high-resolution camera to take real-time motion image capture of the motion process of the moving object, and obtaining real-time motion change image data of the moving object;

[0080] Step S12: Real-time monitoring of the motion acceleration of the motion object corresponding to the motion process is performed by a built-in accelerometer to obtain real-time acceleration data of the motion object;

[0081] Step S13: monitoring the angular velocity of the motion process of the moving body in real time through the built-in gyroscope to obtain the real-time angular velocity data of the moving body;

[0082] Step S14: performing data smoothing preprocessing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body, and the real-time angular velocity data of the moving body to obtain the smoothed data of the motion change image of the moving body, the smoothed data of the acceleration of the moving body, and the smoothed data of the angular velocity of the moving body;

[0083] Step S15: performing time-series alignment and synchronization processing on the moving body motion change image smoothing data, the moving body acceleration smoothing data and the moving body angular velocity smoothing data to obtain the moving body motion image data, the moving body acceleration data and the moving body angular velocity data in the same time-series dimension.

[0084] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in FIG. 1 , in this embodiment, step S1 includes the following steps:

[0085] Step S11: Using a high-resolution camera to take real-time motion image capture of the motion process of the moving object, and obtaining real-time motion change image data of the moving object;

[0086] In an embodiment of the present invention, the movement process of a moving object is captured in real time by using a high-resolution camera. The selected high-resolution camera must have a resolution of at least 1080p and be able to automatically adjust the exposure under different lighting conditions to ensure that clear moving images are captured. The camera is installed at an appropriate position on the moving object to ensure that the camera's field of view can cover the entire movement trajectory of the moving object. After the moving object starts to move, the camera records each frame of the moving object in real time and saves the image data to a local storage device or a cloud server. In order to achieve smooth image capture, the camera's frame rate setting must reach at least 30 frames per second, so that detailed changes can still be captured under fast movement conditions, and finally real-time motion change image data of the moving object is obtained.

[0087] Step S12: Real-time monitoring of the motion acceleration of the motion object corresponding to the motion process is performed by a built-in accelerometer to obtain real-time acceleration data of the motion object;

[0088] In an embodiment of the present invention, real-time acceleration monitoring is performed on the motion process corresponding to the moving body by using a built-in accelerometer. The accelerometer used must have a three-axis measurement function and a high sampling frequency (for example, 100 Hz) so as to capture the acceleration changes of the moving body on the X, Y, and Z axes in real time. When the moving body starts, the accelerometer automatically starts working and obtains the acceleration data of the moving body in different directions through the sensor. These data are transmitted to the data processing unit in real time through a specific communication protocol (such as I2C or SPI), and finally the real-time acceleration data of the moving body is obtained.

[0089] Step S13: monitoring the angular velocity of the motion process of the moving body in real time through the built-in gyroscope to obtain the real-time angular velocity data of the moving body;

[0090] In an embodiment of the present invention, the angular velocity of the motion process corresponding to the moving body is monitored in real time by using a built-in gyroscope. The selected gyroscope needs to be able to measure the angular velocity in three-dimensional space and have high sensitivity and low drift characteristics. After the moving body is started, the gyroscope starts to measure its angular velocity changes in the three axes of X, Y, and Z. Through a specific algorithm, the raw data output by the sensor is filtered and processed to remove noise to ensure accurate angular velocity information. These angular velocity data will be transmitted to the data processing unit in real time and stored with the same timestamp, and finally the real-time angular velocity data of the moving body will be obtained.

[0091] Step S14: performing data smoothing preprocessing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body, and the real-time angular velocity data of the moving body to obtain the smoothed data of the motion change image of the moving body, the smoothed data of the acceleration of the moving body, and the smoothed data of the angular velocity of the moving body;

[0092] In the embodiment of the present invention, the real-time motion change image data, the real-time acceleration data and the real-time angular velocity data of the moving body previously obtained through real-time monitoring are smoothed, and the motion change image data is processed by applying an appropriate smoothing algorithm (such as a moving average method or an exponential smoothing method) to reduce image noise and improve image quality. At the same time, the same smoothing algorithm is also applied to the acceleration data and the angular velocity data to eliminate high-frequency noise and extract more stable trend information. After smoothing, the motion change image data, acceleration data and angular velocity data of the moving body are respectively generated into smoothed data sets, and finally the motion change image smoothed data, the acceleration smoothed data and the angular velocity smoothed data of the moving body are obtained.

[0093] Step S15: performing time-series alignment and synchronization processing on the moving body motion change image smoothing data, the moving body acceleration smoothing data and the moving body angular velocity smoothing data to obtain the moving body motion image data, the moving body acceleration data and the moving body angular velocity data in the same time-series dimension.

[0094] In an embodiment of the present invention, by comparing the timestamps in each data source of the previously smoothed motion change image smoothed data of the moving body, the motion change acceleration smoothed data and the motion change angular velocity smoothed data, it is ensured that the motion change image data, the acceleration data and the angular velocity data of the moving body are compared and analyzed in the same time dimension, and the time difference between the data is identified therefrom, and by applying an interpolation algorithm (such as linear interpolation or spline interpolation) to time-resample the acceleration data and the angular velocity data, so that their timestamps are aligned with the motion change image data. In this way, it is ensured that all data are consistent in the same time dimension, and finally the motion image data, the acceleration data and the angular velocity data of the moving body in the same time series dimension are obtained.

[0095] Further, step S2 includes the following steps:

[0096] Step S21: extracting motion image corner points from each continuous frame of motion image data of the moving object to obtain a set of motion image feature corner points of the moving object;

[0097] Step S22: Establishing a three-dimensional space coordinate system for the moving object motion image data to generate a three-dimensional space coordinate system for the moving object motion image;

[0098] Step S23: performing image space position geometric transformation calculation on each continuous frame of the moving image data of the moving object based on the moving image three-dimensional space coordinate system to obtain the moving image space position geometric registration matrix;

[0099] Step S24: positioning the moving body motion coordinates of each corresponding moving image feature corner point in the moving body motion image feature corner point set based on the moving body motion image spatial position geometric registration matrix to obtain the moving body motion spatial coordinate point set;

[0100] Step S25: Based on the moving body's moving space coordinate point set, long-term and short-term complementary filtering fusion analysis is performed on the moving body's acceleration data and the moving body's angular velocity data to obtain a moving body's moving posture change fusion feature sequence.

[0101] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in FIG. 1 , in this embodiment, step S2 includes the following steps:

[0102] Step S21: extracting motion image corner points from each continuous frame of motion image data of the moving object to obtain a set of motion image feature corner points of the moving object;

[0103] In an embodiment of the present invention, the optical flow method is used in the motion image data of the moving body to extract corner points of each continuous frame of the motion image. The optical flow method identifies motion features by calculating the change in pixel intensity between adjacent frames, and selects pixels with large changes as feature corner points. In a specific implementation, the Shi-Tomasi corner detection algorithm is used to set a threshold to screen corner points with response values ​​higher than the threshold to ensure that the extracted corner points are significant and stable during the motion of the moving body, and finally a feature corner point set of the moving image of the moving body is obtained.

[0104] Step S22: Establishing a three-dimensional space coordinate system for the moving object motion image data to generate a three-dimensional space coordinate system for the moving object motion image;

[0105] In an embodiment of the present invention, a three-dimensional spatial coordinate system is established for the moving image data of the moving object. First, the origin position of the reference coordinate system is determined. Usually, the position of the moving object in a stationary state is selected as the origin. Next, three mutually perpendicular coordinate axes (X, Y, and Z axes) are set to represent the movement of the moving object in different directions. Three-dimensional reconstruction technology is used, and moving images from multiple perspectives are combined. A camera calibration algorithm is used to obtain camera intrinsic and extrinsic parameters to establish a three-dimensional coordinate system. This ensures that the position information of the moving object in each frame in the three-dimensional space can be accurately reflected in the coordinate system, and finally a complete three-dimensional spatial coordinate system for the moving image of the moving object is generated.

[0106] Step S23: performing image space position geometric transformation calculation on each continuous frame of the moving image data of the moving object based on the moving image three-dimensional space coordinate system to obtain the moving image space position geometric registration matrix;

[0107] In an embodiment of the present invention, the image space position geometric transformation calculation is performed on each continuous frame in the moving body moving image data by combining the previously established moving body moving image three-dimensional space coordinate system, so as to convert the two-dimensional image coordinates of each frame into three-dimensional space coordinates by applying the perspective transformation algorithm, and the coordinate data of the moving image is matched with the corresponding points in the three-dimensional space coordinate system by constructing a transformation matrix to obtain a geometric registration matrix. In the registration process, the least squares optimization algorithm is used to minimize the projection transformation parameters between the continuous frames, including position transformation, rotation transformation and scaling transformation parameters, and the transformation parameters are combined into a 4x4 homogeneous transformation matrix. ,Then, the geometric transformation matrix of each frame is used to geometrically align the motion image of the ,moving object, and each pair of adjacent frame matrices is multiplied using linear algebra ,tools (such as the NumPy library), and finally the ,geometric registration matrix of the spatial position of the ,moving image of the moving object is obtained.

[0108] Step S24: positioning the moving body motion coordinates of each corresponding moving image feature corner point in the moving body motion image feature corner point set based on the moving body motion image spatial position geometric registration matrix to obtain the moving body motion spatial coordinate point set;

[0109] In an embodiment of the present invention, after obtaining the geometric registration matrix of the spatial position of the moving image of the moving body, the moving coordinates of the moving body are located for the feature corner point set of the moving image based on the matrix, so that for each feature corner point, its corresponding coordinates in the three-dimensional space coordinate system are calculated using the geometric transformation relationship. In specific operations, the two-dimensional image coordinates of the corner points are transformed in combination with the registration matrix and mapped to the three-dimensional space to generate a moving body moving space coordinate point set. In order to improve the positioning accuracy, the Kalman filtering algorithm is applied to smooth the motion coordinates to reduce the influence of noise, thereby ensuring that the coordinate positioning of the moving body in space is accurate and continuous, and finally the moving body moving space coordinate point set is obtained.

[0110] Step S25: Based on the moving body's moving space coordinate point set, long-term and short-term complementary filtering fusion analysis is performed on the moving body's acceleration data and the moving body's angular velocity data to obtain a moving body's moving posture change fusion feature sequence.

[0111] In an embodiment of the present invention, a long-term and short-term complementary filtering fusion analysis is performed on the acceleration data of the moving body and the angular velocity data of the moving body at the corresponding coordinate points by combining the set of spatial coordinate points of the moving body obtained by the previous registration transformation, so as to smooth the acceleration data by using a low-pass filter to eliminate the influence of high-frequency noise, and at the same time apply a high-pass filter to the angular velocity data to extract its rapidly changing features. Then, a Kalman filter framework is adopted to combine the acceleration and angular velocity data, and through state prediction and updating steps, a characteristic sequence of the movement posture change of the moving body is fused to form. During the analysis process, the time series changes of the fused data are monitored in real time to obtain the posture information of the moving body at different time points, so as to achieve high precision and real-time perception of the movement posture, and finally obtain a fused characteristic sequence of the movement posture change of the moving body.

[0112] Further, step S23 includes the following steps:

[0113] Step S231: generating image coordinate points for each continuous frame of motion image in the motion image data of the moving object, and obtaining a set of motion image coordinate points of the moving object corresponding to each frame of motion image;

[0114] In an embodiment of the present invention, the motion image data of the moving body is segmented into frames to extract the motion image of each frame, and the input continuous motion image sequence is read by using image processing tools, such as the OpenCV library, and then converted into independent image frames by setting the time interval of each frame. In each frame, a feature point detection algorithm (such as SIFT or ORB) is used to identify and extract key feature points. On this basis, image coordinate points are generated, and the extracted feature points are marked in the image coordinate system, and finally a set of motion image coordinate points of the moving body corresponding to each frame of the motion image is obtained.

[0115] Step S232: performing adjacent frame space geometric transformation estimation on the moving body motion image coordinate point set corresponding to each frame of motion image based on the moving body motion image three-dimensional space coordinate system, so as to obtain image space geometric transformation parameters corresponding to each adjacent frame of motion image, wherein the image space geometric transformation parameters include image space position transformation parameters, image space rotation transformation parameters and image space scaling transformation parameters;

[0116] In an embodiment of the present invention, the spatial geometric transformation estimation calculation between adjacent frames is performed on the moving body motion image coordinate point set corresponding to each frame of motion image by combining the previously established moving body motion image three-dimensional spatial coordinate system, so as to randomly select two adjacent frames of images and extract their respective motion image coordinate point sets. At the same time, the motion changes between adjacent frames are estimated by using an optical flow method or a feature point-based matching method (such as a RANSAC algorithm), and the corresponding geometric mathematical relationship between the feature points between the two frames is calculated to obtain the image space position transformation parameters, image space rotation transformation parameters and image space scaling transformation parameters of each frame. In this process, affine transformation or perspective transformation is performed to obtain more accurate transformation parameters, and finally the image space geometric transformation parameters corresponding to each adjacent frame of motion images are obtained.

[0117] Step S233: performing geometric registration matrix transformation calculation according to the image space geometric transformation parameters corresponding to each adjacent frame of motion images to obtain a geometric registration matrix of the motion image space position of the moving object.

[0118] In the embodiment of the present invention, the geometric registration matrix is ​​calculated based on the image space geometric transformation parameters between adjacent frames obtained by the previous analysis. First, the geometric transformation matrix of each pair of adjacent frames is constructed. The position transformation, rotation transformation and scaling transformation parameters can be combined into a 4x4 homogeneous transformation matrix. Then, the geometric transformation matrix of each frame is used to geometrically align the motion image of the moving object, and linear algebra tools (such as the NumPy library) are used to multiply each pair of adjacent frame matrices to obtain the final spatial position geometric alignment matrix. This matrix will be used for subsequent motion posture analysis. By accurately positioning the moving object in three-dimensional space, the comprehensive perception and analysis of the moving body's motion posture can be achieved, and finally the spatial position geometric alignment matrix of the moving image of the moving object can be obtained.

[0119] Further, step S25 includes the following steps:

[0120] Step S251: performing point matching and division on the acceleration data and angular velocity data of the moving body based on each moving space coordinate point in the moving body moving space coordinate point set, and obtaining the moving body acceleration and moving body angular velocity corresponding to each moving space coordinate point;

[0121] In an embodiment of the present invention, by matching the corresponding spatial coordinates of the moving body acceleration data and the moving body angular velocity data in the moving space with each moving body coordinate point in the moving body moving space coordinate point set obtained by combining with the previous positioning, a point motion state data set is formed. Specifically, the acceleration and angular velocity data of the moving body can be obtained by an accelerometer and a gyroscope sensor. Combined with the position of the coordinate point, the dynamic characteristic analysis of the moving body at different points can be realized. Through the interpolation algorithm, the acceleration and angular velocity data can be accurately mapped to each moving space coordinate point, ensuring that each coordinate point can accurately reflect the dynamic characteristics of the moving body at that position, and finally obtaining the moving body acceleration and moving body angular velocity corresponding to each moving space coordinate point.

[0122] Step S252: monitoring the dynamic response duration of the moving body for each moving space coordinate point in the moving body moving space coordinate point set, and obtaining the corresponding motion response duration at each moving space coordinate point;

[0123] In an embodiment of the present invention, by monitoring the dynamic response duration of each motion space coordinate point in the motion space coordinate point set of the motion body, this process can be achieved by setting a timer and synchronizing the sensor data, so as to utilize the data acquisition system to continuously record the timing changes of the motion body at each coordinate point, and calculate the duration of the change. The specific implementation process includes, at each coordinate point, when the motion body is subjected to some external stimulus (such as acceleration, deceleration or turning), recording the start time and end time of the acceleration change, calculating the duration of the change, thereby obtaining the dynamic response duration of the corresponding coordinate point, and finally obtaining the corresponding motion response duration at each motion space coordinate point.

[0124] Step S253: According to the preset response time threshold, the motion response time corresponding to each motion space coordinate point is traversed point by point to determine; when the motion response time is greater than or equal to the preset response time threshold, the corresponding motion space coordinate point is marked as a long-term response coordinate point; when the motion response time is less than the preset response time threshold, the corresponding motion space coordinate point is marked as a short-term response coordinate point;

[0125] In an embodiment of the present invention, a traversal judgment is performed on the response time of each motion space coordinate point by using a preset response time threshold, so as to set a clear response time threshold (for example, 100 milliseconds), and then compare the dynamic response time of each coordinate point in turn. If the response time of a coordinate point is greater than or equal to the threshold, it is marked as a long-term response coordinate point; conversely, if the response time is less than the threshold, it is marked as a short-term response coordinate point. In a specific implementation, a traversal algorithm can be written in a programming language (such as Python or C++), and automatic marking can be achieved by conditional judgment of time series data. This process can effectively screen out coordinate points that have a persistent response to environmental changes.

[0126] Step S254: merging the long-term motion state parameters of the moving body acceleration and the moving body angular velocity corresponding to the long-term response coordinate point marked as the moving body to obtain the moving body long-term response motion state parameter subset; merging the short-term motion state parameters of the moving body acceleration and the moving body angular velocity corresponding to the short-term response coordinate point marked as the moving body to obtain the moving body short-term response motion state parameter subset; performing motion posture change feature analysis on the moving body long-term response motion state parameter subset and the moving body short-term response motion state parameter subset respectively to obtain the moving body long-term motion posture change feature subset and the moving body short-term motion posture change feature subset;

[0127] In an embodiment of the present invention, the acceleration and angular velocity of the moving body at the long-term response and short-term response coordinate points are merged and analyzed respectively, wherein for the coordinate points marked as long-term response, the corresponding acceleration and angular velocity data are collected, and these data are averaged to form a long-term motion state parameter subset, and similar processing is performed on the data of the short-term response coordinate points, and the data are merged by statistical methods (such as weighted averaging) to obtain a short-term motion state parameter subset, and then, feature extraction technology (such as principal component analysis or waveform analysis) is used to perform motion posture change feature analysis on these two subsets to extract the change law of the moving body in different states. The analysis result can provide the motion characteristics of the moving body under long-term and short-term responses, and finally obtain the long-term motion posture change feature subset of the moving body and the short-term motion posture change feature subset of the moving body.

[0128] Step S255: performing long-term and short-term complementary filtering fusion analysis on the long-term motion posture change feature subset of the moving body and the short-term motion posture change feature subset of the moving body to obtain a fusion feature sequence of the motion posture change of the moving body.

[0129] In an embodiment of the present invention, a long-term and short-term complementary filtering method is used to perform fusion analysis on a feature subset of a moving body's long-term motion posture change and a corresponding motion posture feature component in a feature subset of a moving body's short-term motion posture change. This process can be achieved by designing a composite filter that can process long-term and short-term data simultaneously. The specific steps include: first, designing filter parameters to adapt to the smoothness of long-term data and the sensitivity of short-term data; then, applying the filter to fuse the two feature subsets to generate a comprehensive motion posture change feature sequence. The fused feature sequence can effectively reflect the overall motion state of the moving body, and combined with the frequency domain analysis and time domain analysis methods, a comprehensive understanding of the motion posture is provided, and finally a fused feature sequence of the moving body's motion posture change is obtained.

[0130] Further, step S255 includes the following steps:

[0131] Separating the dynamic and static features of the long-term motion posture change feature subset and the short-term motion posture change feature subset of the moving body respectively, and obtaining the dynamic feature component of the long-term motion posture of the moving body, the static feature component of the long-term motion posture of the moving body, the dynamic feature component of the short-term motion posture of the moving body, and the static feature component of the short-term motion posture of the moving body;

[0132] In an embodiment of the present invention, dynamic and static features are separated for a long-term motion posture change feature subset and a short-term motion posture change feature subset of a moving body, respectively, so as to collect posture data of the moving body in different time periods, which data can be obtained by devices such as sensors, accelerometers or gyroscopes, and a method combining time domain analysis and frequency domain analysis is adopted. First, data preprocessing is performed, including denoising and normalization. Then, the long-term motion posture change feature subset is analyzed by a defined feature extraction algorithm to identify dynamic features and static features. The specific operation includes slicing and analyzing the data by using a sliding window technology, and calculating the mean and variance of the data in each window. To judge its dynamic nature, the dynamic feature components are mainly composed of signals with high changing frequency, while the static feature components reflect the posture maintenance state; the short-term motion posture change feature subset also uses the same feature extraction method, but the window size and overlap ratio can be adjusted appropriately to capture more refined short-term change features, thereby obtaining four subsets of dynamic feature components, static feature components of the long-term motion posture of the moving body, and dynamic feature components and static feature components of the short-term motion posture. Finally, the dynamic feature components of the long-term motion posture of the moving body, the static feature components of the long-term motion posture of the moving body, the dynamic feature components of the short-term motion posture of the moving body, and the static feature components of the short-term motion posture of the moving body are obtained.

[0133] Preferably, the characteristic component cumulative error calculation formula is used to calculate the long-term and short-term cumulative errors of the dynamic characteristic component of the long-term motion posture of the moving body, the static characteristic component of the long-term motion posture of the moving body, the dynamic characteristic component of the short-term motion posture of the moving body, and the static characteristic component of the short-term motion posture of the moving body, so as to obtain the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body;

[0134] In an embodiment of the present invention, a suitable characteristic component cumulative error calculation formula is constructed by combining the total observation time of long-term motion posture, the total observation time of short-term motion posture, the integral time variable parameter, the quantitative parameter of the long-term motion posture feature, the quantitative parameter of the short-term motion posture feature, the dynamic characteristic component of the long-term motion posture of the moving body, the static characteristic component of the long-term motion posture of the moving body, the dynamic characteristic component of the short-term motion posture of the moving body, the static characteristic component of the short-term motion posture of the moving body, the error weight coefficient of the corresponding characteristic component, the dynamic characteristic baseline value of the motion posture of the moving body, the static characteristic baseline value of the motion posture of the moving body and related parameters to calculate the long-term and short-term cumulative errors, so as to calculate the cumulative errors between the dynamic characteristic components and the static characteristic components of the long-term and short-term motion posture and the baseline values ​​respectively, and accumulate the errors at each time point through the formula to finally obtain the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body.

[0135] Preferably, based on the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body, a long-term motion posture change feature subset and a short-term motion posture change feature subset of the moving body are subjected to long-term and short-term complementary filtering fusion analysis to obtain a moving body motion posture change fusion feature sequence.

[0136] In an embodiment of the present invention, based on the long-term and short-term posture cumulative errors of the moving body obtained by previous quantitative calculation, complementary filtering fusion analysis is performed between the long-term motion posture change feature subset and the short-term motion posture change feature subset to obtain a fusion feature sequence of the motion posture change of the moving body, and by comparing and judging the specific numerical values ​​between the long-term cumulative error and the short-term cumulative error obtained by previous quantitative calculation, if the long-term cumulative error is greater than or equal to the short-term cumulative error, it means that the cumulative error of the long-term feature component is larger than that of the short-term, and it is necessary to perform corresponding replacement and complementary operations on the long-term feature component to reduce its cumulative error, and then set a corresponding sliding window (for example, 5 seconds) according to the difference between the two, collect the corresponding feature components in the short-term motion posture change feature subset within the corresponding time window, and filter the feature components within its time length through algorithms such as Kalman filtering to obtain the feature components obtained after filtering, and use them for the long-term motion posture change feature subset in the same The corresponding long-term feature components in the sliding window are replaced and compensated for each other to fuse the short-term feature components in each sliding window with the long-term feature components in the same window, and the two are combined using interpolation or weighted average method. The implementation logic of this process is to supplement the deficiencies of long-term posture data through short-term posture filtering data, that is, to eliminate the accumulated errors in the long-term posture data, and form a long-term motion posture feature complementary sequence. Similarly, if the long-term cumulative error is less than the short-term cumulative error, the short-term posture data is supplemented by long-term posture filtering data, that is, to eliminate the accumulated errors in the short-term posture data, and form a short-term motion posture feature complementary sequence. Then, the two feature complementary sequences are linearly combined through weighted coefficients to obtain the final fused feature sequence. The fused feature sequence can effectively reflect the motion posture changes of the moving body on the time axis, provide more comprehensive posture perception information, and finally obtain the motion posture change fused feature sequence of the moving body.

[0137] Furthermore, the characteristic component cumulative error calculation formula is specifically:

[0138] ;

[0139] ;

[0140] In the formula, is the long-term cumulative error of the motion posture of the moving body, is the cumulative error of the short-term motion posture of the moving body, is the total time of long-term motion posture observation, is the total time of short-term motion posture observation, is the integral time variable parameter, is the total number of long-term motion posture features in the long-term motion posture change feature subset of the moving body, is the item index parameter of the long-term motion posture feature, For in time The corresponding dynamic characteristic reference value of the moving body's motion posture, For in time The static characteristic reference value of the motion posture of the moving body corresponding to the position, For the Long-term motion gesture features in time The corresponding long-term motion posture dynamic characteristic components of the moving body, is the error weight coefficient of the long-term dynamic characteristic component, For the Long-term motion gesture features in time The corresponding static characteristic components of the long-term motion posture of the moving body, is the error weight coefficient of the long-term static characteristic component, is the total number of short-term motion posture features in the short-term motion posture change feature subset of the moving body, is the item index parameter of short-term motion posture feature, For the Short-term motion gesture features in time The corresponding short-term motion posture dynamic characteristic components of the moving body, is the error weight coefficient of the short-term dynamic characteristic component, For the Short-term motion gesture features in time The corresponding static characteristic component of the short-term motion posture of the moving body, is the error weight coefficient of the short-term static characteristic component, is the correction coefficient of the cumulative error of the long-term motion posture of the moving body, It is the correction coefficient of the cumulative error of the moving body’s short-term motion posture.

[0141] The present invention obtains a characteristic component cumulative error calculation formula by using a specific mathematical model and after verification, which is used for long-term and short-term cumulative error accounting of the dynamic characteristic component of the long-term motion posture of the moving body, the static characteristic component of the long-term motion posture of the moving body, the dynamic characteristic component of the short-term motion posture of the moving body, and the static characteristic component of the short-term motion posture of the moving body. The characteristic component cumulative error calculation formula can better understand the performance of the moving body at different time scales by separating the long-term and short-term dynamic and static characteristics, and this separation helps to accurately capture different types of motion patterns and posture changes. The formula provides a method for quantifying the cumulative errors of long-term and short-term motion postures, which can effectively reflect the deviation between the performance of the moving body at different stages and the expected deviation, which provides an important basis for subsequent adjustment and optimization of motion strategies. By using time integral variables, the changes in the motion posture are subjected to detailed time series analysis, and the slight changes in the motion process can be captured, thereby improving the sensitivity of dynamic feature evaluation. In addition, by introducing different weight coefficients, the importance of dynamic and static characteristics in error calculation can be adjusted according to specific scenes or requirements, and this flexibility enhances the adaptability of the calculation formula. In addition, the introduction of correction coefficients enables the cumulative error calculation to be adjusted and optimized, and provides a regulation mechanism to correct the cumulative error under specific conditions, thereby improving the accuracy and reliability of the calculation results. Through the comprehensive analysis of long-term and short-term cumulative errors, the complementarity of dynamic and static features can be achieved, thereby generating a more comprehensive motion posture change feature sequence. This fusion analysis can help provide better decision support in complex environments. In summary, the characteristic component cumulative error calculation formula fully considers the long-term motion posture cumulative error of the moving body. , the cumulative error of the short-term motion posture of the moving body , the total time of long-term motion posture observation , total time of short-term motion posture observation , integral time variable parameter , the total number of long-term motion posture features in the long-term motion posture change feature subset of the moving body , item index parameter of long-term motion posture feature , at time The corresponding dynamic characteristic reference value of the moving body's motion posture , at time The static characteristic reference value of the moving body motion posture corresponding to , No. Long-term motion gesture features in time The corresponding long-term motion posture dynamic characteristic components of the moving body , the error weight coefficient of the long-term dynamic characteristic component , No. Long-term motion gesture features in time The static characteristic components of the long-term motion posture of the moving body corresponding to , the error weight coefficient of the long-term static characteristic component , the total number of short-term motion posture features in the short-term motion posture change feature subset of the moving body , item index parameter of short-term motion posture feature , No. Short-term motion gesture features in time The corresponding dynamic characteristic components of the short-term motion posture of the moving body , the error weight coefficient of the short-term dynamic characteristic component , No. Short-term motion gesture features in time The static characteristic components of the short-term motion posture of the moving body corresponding to , the error weight coefficient of the short-term static characteristic component , correction coefficient of the cumulative error of the long-term motion posture of the moving body , correction coefficient of the cumulative error of the short-term motion posture of the moving body , based on the total time of long-term motion posture observation , integral time variable parameter , the total number of long-term motion posture features in the long-term motion posture change feature subset of the moving body , item index parameter of long-term motion posture feature , at time The corresponding dynamic characteristic reference value of the moving body's motion posture , at time The static characteristic reference value of the moving body motion posture corresponding to , No. Long-term motion gesture features in time The corresponding long-term motion posture dynamic characteristic components of the moving body , the error weight coefficient of the long-term dynamic characteristic component , No. Long-term motion gesture features in time The static characteristic components of the long-term motion posture of the moving body corresponding to and the error weight coefficient of the long-term static characteristic component It constitutes a long-term cumulative error of the motion posture of the moving body. The mathematical function calculation relationship of is:

[0142] , and also by observing the total time according to the short-term motion posture , integral time variable parameter , at time The corresponding dynamic characteristic reference value of the moving body's motion posture , at time The static characteristic reference value of the moving body motion posture corresponding to , the total number of short-term motion posture features in the short-term motion posture change feature subset of the moving body , item index parameter of short-term motion posture feature , No. Short-term motion gesture features in time The corresponding dynamic characteristic components of the short-term motion posture of the moving body , the error weight coefficient of the short-term dynamic characteristic component , No. Short-term motion gesture features in time The static characteristic components of the short-term motion posture of the moving body corresponding to and the error weight coefficient of the short-term static characteristic component It constitutes a short-term cumulative error of the motion posture of the moving body. The mathematical function calculation relationship of is:

[0143] This formula can realize the long-term and short-term cumulative error calculation process of the dynamic characteristic component of the long-term motion posture of the moving body, the static characteristic component of the long-term motion posture of the moving body, the dynamic characteristic component of the short-term motion posture of the moving body, and the static characteristic component of the short-term motion posture of the moving body. At the same time, the correction coefficient of the long-term motion posture cumulative error of the moving body is And the correction coefficient of the cumulative error of the short-term motion posture of the moving body The introduction of can be adjusted according to the error conditions that occur during the calculation process, thereby improving the accuracy and applicability of the calculation formula for the cumulative error of the characteristic component.

[0144] Further, the long-term and short-term complementary filtering fusion analysis of the long-term motion posture change feature subset and the short-term motion posture change feature subset of the moving body based on the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body includes the following steps:

[0145] Compare and judge the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body. If the long-term motion posture cumulative error of the moving body is greater than or equal to the short-term motion posture cumulative error of the moving body, set the short-term motion posture feature sliding window according to the difference between the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body, and perform short-term collaborative filtering on the short-term motion posture change feature subset of the moving body based on the short-term motion posture feature sliding window to obtain the corresponding short-term motion posture filtering feature component of the moving body in each short-term sliding window;

[0146] In an embodiment of the present invention, a comparison is made between the specific numerical values ​​of the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body obtained by previous quantitative calculation. If the long-term motion posture cumulative error of the moving body is greater than or equal to the short-term motion posture cumulative error of the moving body, it means that the cumulative error of the long-term motion posture feature component is larger than that of the short-term motion posture, and it is necessary to perform corresponding replacement and complementary operations on the long-term motion posture feature component to reduce its cumulative error. Then, a corresponding short-term motion posture feature sliding window is set according to the difference between the two. In a specific implementation, a fixed time window (for example, 5 seconds) is set according to the difference between the long-term and short-term cumulative errors, so as to collect the motion posture feature components of the moving body of the corresponding duration in the short-term motion posture change feature subset within the corresponding time window, and filter the motion posture feature components of the moving body within the duration through algorithms such as Kalman filtering to obtain the motion posture filter feature components after filtering, and finally extract the corresponding short-term motion posture filter feature components of the moving body within each short-term sliding window.

[0147] Preferably, the short-term motion posture filtering feature component of the moving body corresponding to each short-term sliding window is short-term replaced and complemented with the long-term motion posture feature component of the moving body corresponding to the long-term motion posture change feature subset in the same short-term sliding window to obtain a complementary sequence of long-term motion posture features of the moving body;

[0148] In an embodiment of the present invention, the short-term motion posture filtering feature components corresponding to the moving body in each short-term sliding window obtained after previous filtering are used to replace the long-term motion posture feature components corresponding to the moving body long-term motion posture change feature subset in the same short-term sliding window, so as to fuse the short-term motion posture feature components in each short-term sliding window with the long-term motion posture feature components in the same window, and combine the two by interpolation or weighted averaging. The implementation logic of this process is to supplement the deficiency of long-term posture data by short-term posture filtering data, that is, to eliminate the accumulated errors in the long-term posture data, to form a complementary sequence of long-term motion posture features, to ensure the balance and accuracy of posture data on different time scales, and thus to improve the overall perception effect of the moving body posture, and finally to obtain a complementary sequence of long-term motion posture features of the moving body.

[0149] Preferably, if the cumulative error of the long-term motion posture of the moving body is less than the cumulative error of the short-term motion posture of the moving body, a long-term motion posture feature sliding window is set according to the difference between the cumulative error of the short-term motion posture of the moving body and the cumulative error of the long-term motion posture of the moving body, and a long-term collaborative filtering is performed on the long-term motion posture change feature subset of the moving body based on the long-term motion posture feature sliding window to obtain the corresponding long-term motion posture filtering feature component of the moving body in each long-term sliding window;

[0150] In an embodiment of the present invention, similarly, if the cumulative error of the long-term motion posture of the moving body is smaller than the cumulative error of the short-term motion posture of the moving body, it means that the cumulative error of the short-term motion posture feature component is larger than that of the long-term, and it is necessary to perform corresponding replacement and complementary operations on the long-term motion posture feature component to reduce its cumulative error, and then set the corresponding long-term motion posture feature sliding window according to the difference between the two. In a specific implementation, a fixed time window (for example, 10 seconds) is set according to the difference between the short-term and long-term cumulative errors, and the motion posture feature components of the moving body of the corresponding duration in the long-term motion posture change feature subset of the moving body are collected within the corresponding time window, and the motion posture feature components of the moving body within the duration are filtered through algorithms such as Kalman filtering to obtain the motion posture filter feature components after filtering, and finally extract the corresponding long-term motion posture filter feature components of the moving body in each long-term sliding window.

[0151] Preferably, the long-term motion posture filtering feature component corresponding to the moving body in each long-term sliding window is subjected to long-term replacement and complementation with the short-term motion posture feature component corresponding to the moving body short-term motion posture change feature subset in the same long-term sliding window to obtain a complementary sequence of the moving body short-term motion posture features;

[0152] In an embodiment of the present invention, similarly, the long-term motion posture filter feature components corresponding to the moving body in each long-term sliding window obtained after previous filtering are used to replace the short-term motion posture feature components corresponding to the moving body short-term motion posture change feature subset in the same long-term sliding window, and the feature components are compensated for each other, so that the long-term motion posture feature components in each long-term sliding window are fused with the short-term motion posture feature components in the same window, and the two are combined using interpolation or weighted averaging. The implementation logic of this process is to supplement the deficiency of short-term posture data with long-term posture filter data, that is, to eliminate the accumulated errors in short-term posture data, to form a complementary sequence of short-term motion posture features, to ensure that the posture data on different time scales are balanced and accurate, and finally to obtain a complementary sequence of short-term motion posture features of the moving body.

[0153] Preferably, a dynamic time warping method is used to perform long-term and short-term dynamic programming fusion analysis on the long-term motion posture feature complementary sequence of the moving body and the short-term motion posture feature complementary sequence of the moving body to obtain the motion posture change fusion feature sequence of the moving body.

[0154] In an embodiment of the present invention, a dynamic time warping (DTW) method is used to perform a fusion analysis of long-term and short-term dynamic programming on a long-term motion posture feature complementary sequence of a moving body and a short-term motion posture feature complementary sequence previously obtained after filtering complementation. In a specific implementation, the two groups of sequences are first preprocessed to ensure their alignment on the time axis. Then, the DTW algorithm is applied to calculate the optimal matching path between the two sequences to obtain a comprehensive feature sequence of the moving body's posture at different time scales. This operation effectively integrates the information of short-term rapid changes and long-term stable trends through dynamic programming, thereby forming a comprehensive motion posture change fusion feature sequence, improving the perception accuracy and response speed of the moving body's motion posture, and finally obtaining a motion posture change fusion feature sequence of the moving body.

[0155] Further, step S3 includes the following steps:

[0156] Step S31: performing motion process parameter description analysis on the motion posture change fusion feature sequence of the moving body to obtain an actual parameter description set of the motion process of the moving body, wherein the actual parameter description set of the motion process of the moving body includes actual change parameters of the acceleration of the moving body and actual change parameters of the angular velocity of the moving body;

[0157] In an embodiment of the present invention, a parameterized description analysis of the motion process is performed on the fused feature sequence of the motion posture change of the moving body obtained after filtering and fusion, so that the motion data (such as acceleration, angular velocity, position, etc.) therein are used to calculate the actual change parameters of acceleration and angular velocity by using a numerical integration method, specifically including time integration of the acceleration data to obtain the velocity, and further integration to obtain the displacement. Similarly, after time integration of the angular velocity data, the rotation angle change of the moving body is generated, so as to summarize these calculation results and finally obtain the actual parameterized description set of the motion process of the moving body, which includes the actual change parameters of acceleration and the actual change parameters of angular velocity.

[0158] Step S32: Based on the actual parameterized description set of the motion process of the moving body, the corresponding motion process is subjected to motion key node identification processing to obtain a motion key node set of the moving body motion process, wherein the motion key node set of the moving body motion process includes a starting point, a turning point and an end point;

[0159] In an embodiment of the present invention, the corresponding motion process is processed by identifying key motion nodes in combination with the actual parameterized description set of the motion process of the moving body obtained by previous analysis, so as to determine the motion state of the moving body corresponding to the motion process by analyzing the acceleration and angular velocity change data of the moving body. In this process, the data is classified using the set threshold value to identify the starting point, turning point and end point. The starting point is determined by the sudden change of acceleration and angular velocity, the turning point is determined according to the severity of the change of acceleration and angular velocity, and the end point is determined by the sudden change of acceleration and angular velocity to zero. In a specific implementation, a sliding window algorithm can be used to gradually analyze the time series data and identify the key motion state change points, thereby sorting out the key nodes of the moving body, including all identified starting points, turning points and end points, and finally obtaining a key node set of the motion process of the moving body.

[0160] Step S33: Based on the actual parameterized description set of the motion process of the moving body and the key node set of the motion process of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition analysis to generate the motion posture change trajectory of the moving body.

[0161] In an embodiment of the present invention, the motion posture change trajectory of the corresponding motion process is identified by combining the actual parameterized description set of the motion process of the moving body obtained by previous analysis and the key node set of the motion process of the moving body, so as to utilize the parameterized description data and key nodes obtained in the previous step to extract the state information (such as position, velocity, acceleration, angular velocity, etc.) of the moving body at the key nodes, and use the interpolation algorithm (such as spline interpolation or linear interpolation) to smooth these state information to generate a continuous motion trajectory, analyze the parameter changes of the moving body in different time periods, and accurately describe the posture trajectory of the moving body through data fitting methods (such as polynomial fitting or Bezier curve fitting), which will comprehensively reflect the changes of the moving body in the entire motion process, and finally identify and fit the motion posture change trajectory of the moving body.

[0162] Furthermore, the present invention also provides a moving body motion posture perception system, which is used to execute the moving body motion posture perception method as described above, and the moving body motion posture perception system comprises:

[0163] The motion state parameter analysis module of the moving body is used to take real-time motion images of the motion process corresponding to the moving body through a high-resolution camera to obtain real-time motion change image data of the moving body; to monitor the motion state parameters of the motion process corresponding to the moving body through a built-in accelerometer and gyroscope in real time to obtain real-time acceleration data of the moving body and real-time angular velocity data of the moving body; to perform time-series alignment and synchronization processing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body and the real-time angular velocity data of the moving body, so as to obtain the motion image data of the moving body, the acceleration data of the moving body and the angular velocity data of the moving body in the same time-series dimension;

[0164] The moving body motion feature filtering and fusion module is used to locate the moving body motion coordinates of the moving body motion image data to obtain the moving body motion space coordinate point set; based on the moving body motion space coordinate point set, the moving body acceleration data and the moving body angular velocity data are subjected to long-term and short-term complementary filtering fusion analysis to obtain the moving body motion posture change fusion feature sequence;

[0165] A motion body motion posture trajectory recognition module is used to perform motion posture trajectory recognition and analysis on the corresponding motion process based on the motion body motion posture change fusion feature sequence to generate the motion body motion posture change trajectory;

[0166] The motion state perception and classification module of the moving body is used to perform posture change pattern recognition and analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; based on the motion posture change pattern of the moving body, the corresponding motion posture change trajectory of the moving body is subjected to motion posture perception and classification to obtain the corresponding motion posture category information of the moving body in the current motion process.

[0167] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0168] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for sensing a motion posture of a moving object, characterized in that: The following steps are involved: Step S1: Using a high-resolution camera to take real-time motion image capture of the motion process of the moving object, and obtain real-time motion change image data of the moving object; The motion state parameters of the motion process corresponding to the moving body are monitored in real time through the built-in accelerometer and gyroscope to obtain the real-time acceleration data and the real-time angular velocity data of the moving body; the real-time motion change image data, the real-time acceleration data and the real-time angular velocity data of the moving body are time-series aligned and synchronized to obtain the motion image data, the acceleration data and the angular velocity data of the moving body in the same time-series dimension; Step S2: positioning the motion coordinates of the moving body on the moving image data to obtain a set of motion space coordinate points of the moving body; performing long-term and short-term complementary filtering fusion analysis on the acceleration data and the angular velocity data of the moving body based on the set of motion space coordinate points of the moving body to obtain a fusion feature sequence of the motion posture change of the moving body; wherein step S2 includes the following steps: Step S21: extracting motion image corner points from each continuous frame of motion image data of the moving object to obtain a set of motion image feature corner points of the moving object; Step S22: Establishing a three-dimensional space coordinate system for the moving object motion image data to generate a three-dimensional space coordinate system for the moving object motion image; Step S23: performing image space position geometric transformation calculation on each continuous frame of the moving image data of the moving object based on the moving image three-dimensional space coordinate system to obtain the moving image space position geometric registration matrix; Step S24: positioning the moving body motion coordinates of each corresponding moving image feature corner point in the moving body motion image feature corner point set based on the moving body motion image spatial position geometric registration matrix to obtain the moving body motion spatial coordinate point set; Step S25: Based on the moving body's motion space coordinate point set, long-term and short-term complementary filtering fusion analysis is performed on the moving body's acceleration data and the moving body's angular velocity data to obtain a fusion feature sequence of the moving body's motion posture change; wherein step S25 includes the following steps: Step S251: performing point matching and division on the acceleration data and angular velocity data of the moving body based on each moving space coordinate point in the moving body moving space coordinate point set, and obtaining the moving body acceleration and moving body angular velocity corresponding to each moving space coordinate point; Step S252: monitoring the dynamic response duration of the moving body for each moving space coordinate point in the moving body moving space coordinate point set, and obtaining the corresponding motion response duration at each moving space coordinate point; Step S253: According to the preset response time threshold, the motion response time corresponding to each motion space coordinate point is traversed point by point to determine; when the motion response time is greater than or equal to the preset response time threshold, the corresponding motion space coordinate point is marked as a long-term response coordinate point; when the motion response time is less than the preset response time threshold, the corresponding motion space coordinate point is marked as a short-term response coordinate point; Step S254: merging the long-term motion state parameters of the moving body acceleration and the moving body angular velocity corresponding to the long-term response coordinate point marked as the moving body to obtain the moving body long-term response motion state parameter subset; merging the short-term motion state parameters of the moving body acceleration and the moving body angular velocity corresponding to the short-term response coordinate point marked as the moving body to obtain the moving body short-term response motion state parameter subset; performing motion posture change feature analysis on the moving body long-term response motion state parameter subset and the moving body short-term response motion state parameter subset respectively to obtain the moving body long-term motion posture change feature subset and the moving body short-term motion posture change feature subset; Step S255: performing long-term and short-term complementary filtering fusion analysis on the long-term motion posture change feature subset of the moving body and the short-term motion posture change feature subset of the moving body to obtain the motion posture change fusion feature sequence of the moving body; wherein step S255 includes the following steps: Separating the dynamic and static features of the long-term motion posture change feature subset and the short-term motion posture change feature subset of the moving body respectively, and obtaining the dynamic feature component of the long-term motion posture of the moving body, the static feature component of the long-term motion posture of the moving body, the dynamic feature component of the short-term motion posture of the moving body, and the static feature component of the short-term motion posture of the moving body; The characteristic component cumulative error calculation formula is used to calculate the long-term and short-term cumulative errors of the dynamic characteristic component of the long-term motion posture of the moving body, the static characteristic component of the long-term motion posture of the moving body, the dynamic characteristic component of the short-term motion posture of the moving body, and the static characteristic component of the short-term motion posture of the moving body, so as to obtain the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body; Based on the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body, a long-term and short-term complementary filtering fusion analysis is performed on the long-term motion posture change feature subset of the moving body and the short-term motion posture change feature subset of the moving body to obtain the motion posture change fusion feature sequence of the moving body; Step S3: Based on the fusion feature sequence of the motion posture change of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition and analysis to generate the motion posture change trajectory of the moving body; Step S4: performing posture change pattern recognition analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; performing motion posture perception classification on the corresponding motion posture change trajectory of the moving body based on the motion posture change pattern of the moving body to obtain the corresponding motion posture category information of the moving body in the current motion process.

2. The method for sensing the motion posture of a moving object according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Using a high-resolution camera to take real-time motion image capture of the motion process of the moving object, and obtaining real-time motion change image data of the moving object; Step S12: Real-time monitoring of the motion acceleration of the motion object corresponding to the motion process is performed by a built-in accelerometer to obtain real-time acceleration data of the motion object; Step S13: monitoring the angular velocity of the motion process of the moving body in real time through the built-in gyroscope to obtain the real-time angular velocity data of the moving body; Step S14: performing data smoothing preprocessing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body, and the real-time angular velocity data of the moving body to obtain the smoothed data of the motion change image of the moving body, the smoothed data of the acceleration of the moving body, and the smoothed data of the angular velocity of the moving body; Step S15: performing time-series alignment and synchronization processing on the moving body motion change image smoothing data, the moving body acceleration smoothing data and the moving body angular velocity smoothing data to obtain the moving body motion image data, the moving body acceleration data and the moving body angular velocity data in the same time-series dimension.

3. The method for sensing the motion posture of a moving object according to claim 1, characterized in that: Step S23 includes the following steps: Step S231: generating image coordinate points for each continuous frame of motion image in the motion image data of the moving object, and obtaining a set of motion image coordinate points of the moving object corresponding to each frame of motion image; Step S232: performing adjacent frame space geometric transformation estimation on the moving body motion image coordinate point set corresponding to each frame of motion image based on the moving body motion image three-dimensional space coordinate system, so as to obtain image space geometric transformation parameters corresponding to each adjacent frame of motion image, wherein the image space geometric transformation parameters include image space position transformation parameters, image space rotation transformation parameters and image space scaling transformation parameters; Step S233: performing geometric registration matrix transformation calculation according to the image space geometric transformation parameters corresponding to each adjacent frame of motion images to obtain a geometric registration matrix of the motion image space position of the moving object.

4. The method for sensing the motion posture of a moving object according to claim 1, characterized in that: The characteristic component cumulative error calculation formula is specifically: ； ； In the formula, is the long-term cumulative error of the motion posture of the moving body, is the cumulative error of the short-term motion posture of the moving body, is the total time of long-term motion posture observation, is the total time of short-term motion posture observation, is the integral time variable parameter, is the total number of long-term motion posture features in the long-term motion posture change feature subset of the moving body, is the item index parameter of the long-term motion posture feature, For in time The corresponding dynamic characteristic reference value of the moving body's motion posture, For in time The static characteristic reference value of the motion posture of the moving body corresponding to the position, For the Long-term motion gesture features in time The corresponding long-term motion posture dynamic characteristic components of the moving body, is the error weight coefficient of the long-term dynamic characteristic component, For the Long-term motion gesture features in time The corresponding static characteristic components of the long-term motion posture of the moving body, is the error weight coefficient of the long-term static characteristic component, is the total number of short-term motion posture features in the short-term motion posture change feature subset of the moving body, is the item index parameter of short-term motion posture feature, For the Short-term motion gesture features in time The corresponding short-term motion posture dynamic characteristic components of the moving body, is the error weight coefficient of the short-term dynamic characteristic component, For the Short-term motion gesture features in time The corresponding static characteristic component of the short-term motion posture of the moving body, is the error weight coefficient of the short-term static characteristic component, is the correction coefficient of the cumulative error of the long-term motion posture of the moving body, It is the correction coefficient of the cumulative error of the moving body’s short-term motion posture.

5. The method for sensing the motion posture of a moving object according to claim 1, characterized in that: The method of performing long-term and short-term complementary filtering fusion analysis on a long-term motion posture change feature subset and a short-term motion posture change feature subset of a moving body based on the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body comprises the following steps: Compare and judge the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body. If the long-term motion posture cumulative error of the moving body is greater than or equal to the short-term motion posture cumulative error of the moving body, set the short-term motion posture feature sliding window according to the difference between the long-term motion posture cumulative error of the moving body and the short-term motion posture cumulative error of the moving body, and perform short-term collaborative filtering on the short-term motion posture change feature subset of the moving body based on the short-term motion posture feature sliding window to obtain the corresponding short-term motion posture filtering feature component of the moving body in each short-term sliding window; The short-term motion posture filtering feature component of the moving body corresponding to each short-term sliding window is short-term replaced and complemented with the long-term motion posture feature component of the moving body corresponding to the long-term motion posture change feature subset in the same short-term sliding window to obtain a complementary sequence of long-term motion posture features of the moving body; If the long-term motion posture cumulative error of the moving body is less than the short-term motion posture cumulative error of the moving body, a long-term motion posture feature sliding window is set according to the difference between the short-term motion posture cumulative error of the moving body and the long-term motion posture cumulative error of the moving body, and a long-term collaborative filtering is performed on the long-term motion posture change feature subset of the moving body based on the long-term motion posture feature sliding window to obtain the corresponding long-term motion posture filtering feature component of the moving body in each long-term sliding window; The long-term motion posture filtering feature component of the moving body corresponding to each long-term sliding window is subjected to long-term replacement and complementation with the short-term motion posture feature component of the moving body corresponding to the short-term motion posture change feature subset in the same long-term sliding window to obtain a complementary sequence of the short-term motion posture features of the moving body; The dynamic time warping method is used to perform long-term and short-term dynamic planning fusion analysis on the long-term motion posture feature complementary sequence and the short-term motion posture feature complementary sequence of the moving body, and the fusion feature sequence of the moving body's motion posture change is obtained.

6. The method for sensing the motion posture of a moving object according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing motion process parameter description analysis on the motion posture change fusion feature sequence of the moving body to obtain an actual parameter description set of the motion process of the moving body, wherein the actual parameter description set of the motion process of the moving body includes actual change parameters of the acceleration of the moving body and actual change parameters of the angular velocity of the moving body; Step S32: Based on the actual parameterized description set of the motion process of the moving body, the corresponding motion process is subjected to motion key node identification processing to obtain a motion key node set of the moving body motion process, wherein the motion key node set of the moving body motion process includes a starting point, a turning point and an end point; Step S33: Based on the actual parameterized description set of the motion process of the moving body and the key node set of the motion process of the moving body, the corresponding motion process is subjected to motion posture trajectory recognition analysis to generate the motion posture change trajectory of the moving body.

7. A system for sensing the motion posture of a moving object, characterized in that: Used to execute the method for sensing the motion posture of a moving body as claimed in claim 1, the system for sensing the motion posture of a moving body comprises: The motion state parameter analysis module of the moving body is used to take real-time motion images of the motion process corresponding to the moving body through a high-resolution camera to obtain real-time motion change image data of the moving body; to monitor the motion state parameters of the motion process corresponding to the moving body through a built-in accelerometer and gyroscope in real time to obtain real-time acceleration data of the moving body and real-time angular velocity data of the moving body; to perform time-series alignment and synchronization processing on the real-time motion change image data of the moving body, the real-time acceleration data of the moving body and the real-time angular velocity data of the moving body, so as to obtain the motion image data of the moving body, the acceleration data of the moving body and the angular velocity data of the moving body in the same time-series dimension; The moving body motion feature filtering and fusion module is used to locate the moving body motion coordinates of the moving body motion image data to obtain the moving body motion space coordinate point set; based on the moving body motion space coordinate point set, the moving body acceleration data and the moving body angular velocity data are subjected to long-term and short-term complementary filtering fusion analysis to obtain the moving body motion posture change fusion feature sequence; A motion body motion posture trajectory recognition module is used to perform motion posture trajectory recognition and analysis on the corresponding motion process based on the motion body motion posture change fusion feature sequence to generate the motion body motion posture change trajectory; The motion state perception and classification module of the moving body is used to perform posture change pattern recognition and analysis on the motion posture change trajectory of the moving body to obtain the motion posture change pattern of the moving body; based on the motion posture change pattern of the moving body, the corresponding motion posture change trajectory of the moving body is subjected to motion posture perception and classification to obtain the corresponding motion posture category information of the moving body in the current motion process.

Images