Motion data processing method and device based on computer vision

Abstract

The invention discloses a motion data processing method and device based on computer vision, and relates to the technical field of motion data processing. A sensing coordinate system, a personal dynamic coordinate system and a coordinate transformation operation method are established, and the motion data coupling phenomenon is optimized. The motion mode of each part of the body of the athlete in the competition process can be accurately reflected; according to the method, a plurality of camera positions are arranged, and each camera position is optimized through numerical value limiting calculation, so that the reasonability of data source arrangement is optimized; the influence of various factors such as equipment precision, environmental factors and manual operation on data acquisition in the data acquisition process is avoided, and the accuracy and reliability of data are ensured; the acquired relative coordinates of each body structure of the athlete are subjected to numerical analysis from five angles of linear velocity, angular velocity, acceleration, amplitude and frequency of movement, and numerical analysis results exceeding a threshold value are highlighted, so that the diversity of a data analysis method is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motion data processing, and in particular to a motion data processing method and device based on computer vision. Background Art

[0002] Sports data processing technology is a complex process involving multiple steps and methods aimed at extracting valuable information from raw data from athletes or sports equipment to optimize training, improve sports performance and reduce the risk of injury.

[0003] Motion data processing technology based on computer vision is an important topic in the field of computer vision. Detecting and tracking moving targets in video sequences and obtaining the target's position, speed and other motion parameters are the basis for more advanced motion target behavior analysis. Rapid and real-time detection of moving targets, tracking of moving targets, and analysis of target motion data can be widely used in the field of sports. For example, a training video of a track athlete is processed to extract information such as its instantaneous motion speed, arm swing amplitude, frequency, etc. Since the video sequence has multiple images per second, it contains a large amount of valuable information that cannot be captured by the human eye, which can be used to analyze the technical deficiencies of athletes and improve sports skills.

[0004] However, the existing motion data processing technology still has many shortcomings and limitations. First, the existing computer vision motion data processing technology has the phenomenon of data coupling. For example, for a track and field athlete who is doing a long-distance running event, if a spatial coordinate system is established for data collection, only the relative movement speed of each part of the athlete's body on the track can be observed. It cannot accurately reflect the movement patterns of the athlete's legs, hips and other parts of the body during the competition, especially in different stages such as acceleration, speed maintenance and deceleration.

[0005] Secondly, the existing sports data processing technology makes it difficult to interpret the results of sports data analysis. Due to the complexity of data analysis technology, many non-professional users find it difficult to accurately interpret the analysis results, resulting in the inability to fully utilize the value of the data. For example, the existing sports data processing technology makes it difficult to comprehensively consider the dynamic changes of the athlete's posture and center of gravity and their impact on the athlete.

[0006] Therefore, it is necessary to design a motion data processing method and device based on computer vision to solve the above problems. Summary of the invention

[0007] In order to overcome the problems in the background technology, the embodiments of the present invention provide a motion data processing method and device based on computer vision.

[0008] The purpose of the present invention can be achieved through the following technical solutions: In a first aspect, the present invention provides a motion data processing method based on computer vision, which specifically comprises the following steps: Step 1: Perception network layout: Arrange several cameras with fixed positions, set up industrial-grade high-definition camera 1 at preset position 1, industrial-grade high-definition camera 2 at preset position 2, and industrial-grade high-definition camera 3 at preset position 3. Let preset position 1 be the coordinate origin O1, establish a three-dimensional rectangular coordinate system with the north direction as the Y axis, the east direction as the Y axis, and the vertical direction as the Z axis, and record it as the perception coordinate system. Obtain the coordinates O2 (Xo2, Yo2, Zo2) and O3 (Xo3, Yo3, Zo3) of preset positions such as preset position 2 and preset position 3 in the perception coordinate system.

[0009] For the preset positions 1, 2 and 3, the preset distance thresholds dmax and dmin, and the preset maximum angles θmax and θmin, numerical limit calculations are performed on the arrangement of the preset positions 1, 2 and 3, specifically: Connect points O1, O2 and O3 in sequence to form a triangle. Let the coordinates of point O1 be (Xo1, Yo1, Zo1), and use the formula The distance a between the preset positions 1 and 2, the distance b between the preset positions 1 and 3, and the distance c between the preset positions 2 and 3 are calculated.

[0010] By distance constraint The distances a, b and c are defined.

[0011] By formula The angles θa, θb and θc of the triangle formed by the preset positions 1, 2 and 3 are calculated.

[0012] By Angle Constraint The angles θa, θb and θc are defined.

[0013] Furthermore, when the number of preset positions exceeds three, that is, when the number of industrial-grade high-definition cameras installed is greater than three, the adjacent preset position coordinate points are sequentially connected to form a plurality of triangles, and the above-mentioned numerical limit calculation is performed for each triangle.

[0014] Adjust each industrial-grade high-definition camera so that its line of sight is facing the sports field. Then, take the athlete's face image and retrieve the athlete's face recognition feature map from the database, perform face recognition comparison based on computer vision technology, match the specific athlete according to the face recognition comparison result and assign a unique number.

[0015] The unique number is i, i=1, 2, ..., n; wherein n is the total number of athletes included in the system.

[0016] Step 2: image processing, joint recognition and human body model construction; The motion images of each athlete at each time t are obtained by recording and photographing with industrial-grade high-definition cameras and decomposing them frame by frame. Distortion correction and image noise reduction are performed on the motion images.

[0017] As a preferred embodiment of the present invention, the motion images are feature extracted based on computer vision technology to construct a three-dimensional digital twin model of each athlete, and the body structures of the athletes are located in the three-dimensional digital twin model through a key point detection neural network model. The body structure is divided into a connection part, a backbone part and a limb part according to the motion mode, and they are numbered as follows: Connection parts: A01 neck joint, A02 left shoulder joint, AO3 right shoulder joint, AO4 waist joint, A05 left hip joint, A06 right hip joint, A07 left elbow joint, A08 left wrist joint, A09 right elbow joint, A10 right wrist joint, A11 left knee joint, A12 right knee joint, A13 left ankle joint and A14 right ankle joint; The backbone: B01 head, B02 upper torso, B03 lower torso, B04 left palm, B05 right palm, B06 left sole and B07 right sole; Limb parts: C01 left upper arm, C02 right upper arm, C03 left forearm, C04 right forearm, C05 left thigh, C06 right thigh, C07 left calf and C08 right calf.

[0018] As a preferred embodiment of the present invention, geometric abstraction is performed on the three-dimensional digital twin model of each athlete, and the specific form of the human body is abstracted into a combination of a series of geometric figures, thereby generating a movement diagram.

[0019] The geometric abstraction process is specifically as follows: using circles to simplify the connection parts, using polygons to simplify the backbone parts, and using line segments to simplify the limb parts.

[0020] Furthermore, at every preset time interval t, characteristic coordinate points are extracted from the motion diagram to quantify the motion data, and the specific process is as follows: Locate the geometric centroid of the entire motion diagram of each athlete, that is, the geometric center of the entire image composed of all circles and polygons in each motion diagram. Record the coordinates (xi, yi, zi) of the geometric centroid of each athlete's motion diagram in the perception coordinate system at every preset time interval t, and record them as positioning coordinates. For each athlete's motion diagram, in the perception coordinate system, obtain the positioning coordinates of each body structure at every preset time interval t, specifically: For the connecting part, the coordinates of the center of each circle are used as the positioning coordinates of the connecting part; for the backbone part, the coordinates of the center of each polygon are used as the positioning coordinates of the backbone part; for the limb part, the coordinates of the two endpoints of each line segment are used as the positioning coordinates of the limb part.

[0021] Step 3: Coordinate transformation: Establish a personal dynamic coordinate system, specifically: As a preferred embodiment of the present invention, for each athlete's motion diagram, locate the polygon corresponding to the upper torso, obtain the geometric center coordinates iB02 (XiB02, YiB02, ZiB02) of the polygon, and record the plane where the polygon is located as the personal dynamic reference plane.

[0022] Draw a straight line in space, and make the straight line meet the following conditions: Condition 1, perpendicular to the personal dynamic reference plane; Condition 2, the straight line passes through the geometric center of the polygon; The straight line is recorded as reference normal vector one.

[0023] Furthermore, in the polygon corresponding to the upper torso, obtain the corner point closest to the head positioning coordinate iB01 (XiB01, YiB01, ZiB01), which is recorded as the feature point. Obtain the coordinates D (Xia, Yib, Zic) of the feature point. Draw a spatial straight line that satisfies the following conditions: Condition 1: the straight line is located in the personal dynamic reference plane; Condition 2: the straight line passes through the geometric center coordinates of the polygon; Condition 3: the straight line passes through the feature point.

[0024] The straight line is recorded as reference normal vector two.

[0025] A straight line perpendicular to the reference normal vector 2 is drawn in the personal dynamic reference plane, and is named reference normal vector 3. The reference normal vectors 1, 2 and 3 are perpendicular to each other.

[0026] With the centroid coordinates of each athlete as the coordinate base point, the reference normal vector 1 as the X-axis, the reference normal vector 2 as the Y-axis, and the reference normal vector 3 as the Z-axis, a spatial rectangular coordinate system is established, which is recorded as the personal dynamic coordinates of athlete i.

[0027] Where i is the unique number of the athlete, and t is the current moment. In the personal dynamic reference plane.

[0028] Furthermore, a coordinate transformation operation relationship is established between the perception coordinate system and the personal dynamic coordinate system, specifically: Get the coordinates of the geometric centroid of each athlete (Xi, Yi, Zi), and let the translation transformation matrix be: , get the angle θ1 between the reference normal vector 1 and the X axis in the perception coordinate system, get the angle θ2 between the reference normal vector 2 and the Y axis in the perception coordinate system, get the angle θ3 between the reference normal vector 3 and the Z axis in the perception coordinate system. Define the rotation transformation matrix as: .

[0029] Let the composite transformation matrix M be Given a point P (x, y, z) in the perception coordinate system, through the composite transformation matrix M, the new coordinates P ´ (x ´, y ´, z ´) of the point in the personal dynamic coordinate system can be calculated as follows: .

[0030] Furthermore, for each athlete, the collected positioning coordinates of each body structure of the athlete are transformed into the athlete's personal dynamic coordinate system in real time through the coordinate transformation operation relationship to obtain the relative coordinates of each body structure, including: The relative coordinates of the connecting part, the relative coordinates of the backbone part and the relative coordinates of the limb part.

[0031] Step 4: Numerical analysis For each athlete i, all body structures are uniquely numbered, and all connecting parts are uniquely numbered.

[0032] According to the relative coordinates of each body structure, the specific parts of all body structures are uniquely numbered, and the unique number is j1, j1=1, 2, ..., m1; wherein m is the total number of body parts.

[0033] Each connection part in the body structure is uniquely numbered, and the unique number is j2, j2=1, 2, ..., m2; wherein m2 is the total number of connection parts.

[0034] The relative coordinates of each body structure of the athlete collected are numerically analyzed from five perspectives: linear velocity, angular velocity, acceleration, amplitude and frequency of movement. Specifically: Motion linear velocity analysis: For each preset time interval t, the relative coordinates of each body structure are recorded. The relative coordinates of the current time E1 = (XE1, YE1, ZE1) are recorded, and the relative coordinates of the previous preset time interval t corresponding to the same body structure E2 = (XE2, YE2, ZE2) are obtained. By formula Calculate the linear displacement S(i1, j, T) of athlete i's body structure j at time T. Calculate the linear velocity VS(i, j1, T) of athlete i's body structure j1 at time T by the formula VS=S(i, j, T) / t; Analysis of angular velocity of motion: Obtain the self-relative coordinates of the connected parts. For each connected part, obtain the self-relative coordinates of the limb part and the backbone part connected to it in the motion diagram to form a connecting polyline. For the two self-relative coordinates of the end points of the limb part line segment, select the end point of the line segment according to the principle of the distal end point, and form a connecting polyline with the connected part.

[0035] At every preset time interval t, the angles of each connecting polyline are obtained through spatial coordinate calculation. The specific process is as follows: In the connecting polyline, let the two end points be F1 and F2, and let the middle connecting point be F0.

[0036] Get their coordinates F0 (XF0, YF0, ZF0), F1 (XF1, YF1, ZF1) and F3 (XF3, YF3, ZF3) respectively. Calculate the connection line angle θ(i, j2, T) corresponding to the connection part j2 of athlete i at the current moment, where i is the unique number of each athlete; j2 is the unique number of each connection part; and T is the current moment. Get the previous preset time interval t and calculate the connection line angle θ(i, j2, Tt) corresponding to the connection part j2 of athlete i.

[0037] The changing speed of the angle θ(i, j2, T) of the connecting broken line is calculated by the formula vθ(i, j2, T)=θ[(i, j2, T)-θ(i, j2, Tt)] / t, which is recorded as the angular velocity of the angle of each connecting part.

[0038] Motion acceleration analysis: Based on the calculation results of the motion linear velocity analysis, the motion linear velocity VS (i, j1, T) of the athlete i's body structure j1 at time T and the motion linear velocity VS´ (i, j1, Tt) at the previous preset time interval t are obtained, and the linear acceleration of the athlete i's body structure j1 at time T is calculated by the formula aS (i, j2, T) = [VS (i, j1, T) - VS´ (i, j1, Tt)] / t.

[0039] Based on the calculation results of the motion angular velocity analysis, the motion angular velocity Vθ(i, j2, T) of the connecting part j2 of athlete i at time T and the motion linear velocity Vθ´(i, j2, Tt) at the previous preset time interval t are obtained, and the linear acceleration of the connecting part j2 of athlete i at time T is calculated by the formula aθ(i, j2, T)=[Vθ(i, j2, T)-Vθ´(i, j2, Tt)] / t.

[0040] Motion amplitude analysis: preset an amplitude monitoring period T1=n1×t, where n1 is a preset period base, n1 is a positive integer greater than 0; and t is a preset time interval.

[0041] Based on the analysis and calculation results of the linear velocity of motion, every amplitude monitoring period T1, the linear displacement of the body structure j of athlete i at each moment is obtained and its maximum value Smax(i1, j, T) is extracted, which is recorded as the motion amplitude of the body structure j of athlete i within the amplitude monitoring period T1.

[0042] Motion frequency analysis: Preset a frequency monitoring period T2=n2×t, where n2 is the preset frequency base, n2 is a positive integer greater than 0; and t is the preset time interval.

[0043] A minimum displacement benchmark Smin is preset. Based on the analysis and calculation results of the linear velocity of the movement, the linear displacement of the body structure j of the athlete i at each moment is obtained. Subsequently, the linear displacement greater than the minimum displacement benchmark Smin is marked as a valid linear displacement. Every amplitude monitoring period T1, the number of valid linear displacements Q is counted and recorded as the movement frequency.

[0044] The analysis and calculation of the linear velocity, angular velocity, acceleration, amplitude and frequency are repeatedly performed at preset time intervals t, so that the analysis and calculation process traverses the unique numbers i of all athletes and the unique numbers j1 of all body structures.

[0045] Step 5: Visual display of numerical values; The data visualization of the motion diagram is as follows: The specific values ​​of the linear velocity, angular velocity, linear acceleration, angular acceleration, amplitude and frequency of movement are marked in the movement diagram according to the unique number i of the athlete and the unique number j1 of all body structures.

[0046] A set of data thresholds are preset, including a motion linear velocity threshold, a motion angular velocity threshold, a linear acceleration threshold, an angular acceleration threshold, a motion amplitude threshold, and a motion frequency threshold. Specific values ​​exceeding their own thresholds are highlighted in the marked motion diagram.

[0047] Furthermore, the motion diagrams processed by data visualization are separated and saved at time intervals t, providing a playback window for the motion diagrams over time.

[0048] Step 6: Data preservation; The positioning coordinates of all body structures of each athlete, the positioning coordinates after coordinate conversion, the linear velocity, angular velocity, acceleration, amplitude and frequency of movement obtained through analysis and calculation are saved.

[0049] Furthermore, the motion diagram with specific numerical annotations and highlights is saved.

[0050] In a second aspect, the present invention provides a motion data processing device based on computer vision, which is used for video photography, and includes a tripod, a circular support platform is installed on the upper end of the tripod, and an industrial-grade high-definition camera for shooting is installed on the upper end of the support platform. The lower end of the tripod is provided with a freely extendable support leg 1, a support leg 2 and a support leg 3. The support leg 1, the support leg 2 and the support leg 3 intersect at the upper end of the tripod and are rigidly connected by a triangular horizontal limit plate, and a limit hole is provided at the top of the tripod.

[0051] The support platform includes a coupling component 1 at the upper end and a circular support platform plane at the lower end. The lower end of the support platform plane is limitedly connected to the limit hole in the tripod to limit the displacement of the support platform and ensure that the support platform plane rotates with the limit hole as the center of the circle during use. The coupling component 1 forms a limited connection between the support platform plane and the camera pan / tilt in the industrial-grade high-definition camera to limit the displacement of the industrial-grade high-definition camera and ensure that the industrial-grade high-definition camera rotates with the coupling component 1 as the center of the circle during use.

[0052] The industrial-grade high-definition camera includes a camera pan platform at the lower end. The camera pan platform is connected to the camera body in a limited position through a coupling component 2 on the side of the camera body.

[0053] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention establishes a perception coordinate system and a personal dynamic coordinate system, and proposes a coordinate transformation calculation method, which optimizes the phenomenon of motion data coupling; accurately reflects the movement patterns of athletes' legs, hips and other body parts during the game, especially in different stages such as acceleration, speed maintenance and deceleration; (2) The present invention sets multiple preset positions and multiple camera positions, and optimizes each camera position through numerical limit calculation, thereby optimizing the rationality of data source layout; avoiding the influence of multiple factors such as equipment accuracy, environmental factors, and human operation on data collection during data collection, and ensuring the accuracy and reliability of data; (3) The present invention performs numerical analysis on the relative coordinates of each body structure of the athlete collected from five perspectives: linear velocity, angular velocity, acceleration, amplitude and frequency of the movement, and highlights the numerical analysis results that exceed the threshold, thereby improving the diversity of data analysis methods; establishing a simple movement diagram and performing numerical analysis on each body part of the athlete, which is conducive to users to dig out hidden values ​​and rules from massive data; at the same time, the difficulty of data interpretation is reduced through data visualization processing; performing targeted numerical analysis on each body part of the athlete, establishing a big data mindset, and making full use of the value of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings: Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a schematic diagram of the three-camera arrangement of the present invention; Figure 3 It is a schematic diagram of the multi-camera arrangement of the present invention; Figure 4 A schematic diagram of a three-dimensional digital twin model of an athlete of the present invention; Figure 5 A schematic diagram of a sportsman's movement diagram of the present invention; Figure 6 A schematic diagram of a personal dynamic coordinate system of the present invention; Figure 7 It is a schematic diagram of the connection line of the present invention; Figure 8 It is a structural schematic diagram of a motion data processing device based on computer vision; Fig. 9 It is a structural explosion diagram of a motion data processing device based on computer vision.

[0055] Figure numerals: 1, tripod; 2, support platform; 3, industrial-grade high-definition camera; 101, support leg one; 102, support leg two; 103, support leg three; 104, horizontal limit plate; 105, limit hole; 201, support platform plane; 202, coupling component one; 301, camera pan head; 302, camera body; 303, coupling component two. DETAILED DESCRIPTION

[0056] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1 As shown, the motion data processing method based on computer vision specifically includes the following steps: Step 1: Perception network layout: See also Figure 2As shown, three cameras with fixed positions are arranged. Industrial HD camera 1 is set up at preset position 1, industrial HD camera 2 is set up at preset position 2, and industrial HD camera 3 is set up at preset position 3. Let preset position 1 be the coordinate origin O1, and establish a three-dimensional rectangular coordinate system with the north direction as the Y axis, the east direction as the Y axis, and the vertical direction as the Z axis, which is recorded as the perception coordinate system. Obtain the coordinates O2 (Xo2, Yo2, Zo2) and 03 (Xo3, Yo3, Zo3) of preset position 2 and preset position 3 in the perception coordinate system respectively.

[0058] The preset distance thresholds dmax and dmin, the preset maximum angles θmax and θmin are used to perform numerical calculations on the arrangement of the preset positions one, two and three, specifically: Connect points O1, O2 and O3 in sequence to form a triangle. Let the coordinates of point O1 be (Xo1, Yo1, Zo1), and use the formula The distance a between the preset positions 1 and 2, the distance b between the preset positions 1 and 3, and the distance c between the preset positions 2 and 3 are calculated.

[0059] By distance constraint The distances a, b and c are defined.

[0060] By formula The angles θa, θb and θc of the triangle formed by the preset positions 1, 2 and 3 are calculated.

[0061] By Angle Constraint The angles θa, θb and θc are defined.

[0062] For further information, see Figure 3 As shown, if more than three preset positions are set, that is, more than three industrial-grade high-definition cameras are installed, the adjacent preset position coordinate points are sequentially connected to form a number of triangles. The above numerical limit calculation is performed for each triangle.

[0063] Adjust each industrial-grade high-definition camera so that its line of sight is facing the sports field. Then, take the athlete's face image and retrieve the athlete's face recognition feature map from the database, perform face recognition comparison based on computer vision technology, match the specific athlete according to the face recognition comparison result and assign a unique number.

[0064] The unique number is i, i=1, 2, ..., n; wherein n is the total number of athletes included in the system.

[0065] Step 2: image processing, joint recognition and human body model construction; The motion images of each athlete at each time t are obtained by recording and photographing with industrial-grade high-definition cameras and decomposing them frame by frame. Distortion correction and image noise reduction are performed on the motion images.

[0066] See also Figure 4 As shown, the sports image is feature extracted based on computer vision technology to construct a three-dimensional digital twin model of each athlete, and the body structures of the athlete are located in the three-dimensional digital twin model through a key point detection neural network model. The body structure is divided into a connection part, a backbone part and a limb part according to the movement mode, and they are numbered as follows: Connection parts: A01 neck joint, A02 left shoulder joint, AO3 right shoulder joint, AO4 waist joint, A05 left hip joint, A06 right hip joint, A07 left elbow joint, A08 left wrist joint, A09 right elbow joint, A10 right wrist joint, A11 left knee joint, A12 right knee joint, A13 left ankle joint and A14 right ankle joint; The backbone: B01 head, B02 upper torso, B03 lower torso, B04 left palm, B05 right palm, B06 left sole and B07 right sole; Limb parts: C01 left upper arm, C02 right upper arm, C03 left forearm, C04 right forearm, C05 left thigh, C06 right thigh, C07 left calf and C08 right calf.

[0067] See also Figure 5 As shown, geometric abstraction is performed on the three-dimensional digital twin model of each athlete, and the specific form of the human body is abstracted into a combination of a series of geometric figures to generate a simple motion diagram.

[0068] The geometric abstraction process is specifically as follows: using circles to simplify the connection parts, using polygons to simplify the backbone parts, and using line segments to simplify the limb parts.

[0069] Furthermore, at every preset time interval t, characteristic coordinate points are extracted from the motion diagram to quantify the motion data, and the specific process is as follows: Locate the geometric centroid of the entire motion diagram of each athlete, that is, the geometric center of the entire image composed of all circles and polygons in each motion diagram. Record the coordinates (xi, yi, zi) of the geometric centroid of each athlete's motion diagram in the perception coordinate system at every preset time interval t, and record them as positioning coordinates. For each athlete's motion diagram, in the perception coordinate system, obtain the positioning coordinates of each body structure at every preset time interval t, specifically: For the connection part, the center coordinates of each circle are used as the positioning coordinates of the connection part, including: neck joint positioning coordinates iA01 (XiA01, YiA01, ZiA01), left shoulder joint positioning coordinates iA02 (XiA02, YiA02, ZiA02), right shoulder joint positioning coordinates iA03 (XiA03, YiA03, ZiA03), waist joint positioning coordinates iA04 (XiA04, YiA04, ZiA04), left hip joint positioning coordinates iA05 (XiA05, YiA05, ZiA05), right hip joint positioning coordinates iA06 (XiA06, YiA06, ZiA06), left elbow joint positioning coordinates iA07 (XiA08, YiA08, ZiA08), right elbow joint positioning coordinates iA09 (XiA01, YiA01, ZiA01), right flexor joint positioning coordinates iA100 (XiA11, YiA12, ZiA13), right flexor joint positioning coordinates iA111 (XiA12, YiA12, ZiA13), right flexor joint positioning coordinates iA132 (XiA13, YiA13, ZiA13), left flexor joint positioning coordinates iA14 (XiA14, YiA14, ZiA14), right flexor joint positioning coordinates iA15 (XiA15, YiA15, ZiA15), left flexor joint positioning coordinates iA16 (XiA16, YiA16, ZiA16), right flexor joint positioning coordinates iA17 (XiA17 7, YiA07, ZiA07), left wrist joint positioning coordinates iA08 (XiA08, YiA08, ZiA08), right elbow joint positioning coordinates iA09 (XiA09, YiA09, ZiA09), right wrist joint positioning coordinates iA10 (XiA10, YiA10, ZiA10), left knee joint positioning coordinates iA11 (XiA11, YiA11, ZiA11), right knee joint positioning coordinates iA12 (XiA12, YiA12, ZiA12), left ankle joint positioning coordinates iA13 (XiA13, YiA13, ZiA13) and right ankle joint positioning coordinates iA14 (XiA14, YiA14, ZiA14); For the backbone part, the center coordinates of each polygon are used as the positioning coordinates of the backbone part, including: head positioning coordinates iB01 (XiB01, YiB01, ZiB01), upper trunk positioning coordinates iB02 (XiB02, YiB02, ZiB02), lower trunk positioning coordinates (XiB03, YiB03, ZiB03), left palm positioning coordinates iB04 (XiB04, YiB04, ZiB04), right palm positioning coordinates iB05 (XiB05, YiB05, ZiB05), left foot positioning coordinates iB06 (XiB06, YiB06, ZiB06) and right foot positioning coordinates iB07 (XiB07, YiB07, ZiB07); For the limbs, the two endpoints of each line segment are used as the positioning coordinates of the limbs. Including: left upper arm positioning coordinates iC01-1 (XiC01-1, YiC01-1, ZiC01-1) and iC01-2 (XiC01-2, YiC01-2, ZiC01-2), right upper arm positioning coordinates iC02-1 (XiC02-1, YiC02-1, ZiC02-1) and iC02-2 (XiC02-2, YiC02-2, ZiC02-2 )、Left forearm positioning coordinates iC03-1 (XiC03-1, YiC03-1, ZiC03-1) and iC03-2 (XiC03-2, YiC03-2, ZiC03-2), right forearm positioning coordinates iC04-1 (XiC04-1, YiC04-1, ZiC04-1) and iC04-2 (XiC04-2, YiC04-2, ZiC04-2) , left thigh positioning coordinates iC05-1 (XiC05-1, YiC05-1, ZiC05-1) and iC05-2 (XiC05-2, YiC05-2, ZiC05-2), right thigh positioning coordinates iC06-1 (XiC06-1, YiC06-1, ZiC06-1) and iC06-2 (XiC06-2, YiC06-2, ZiC06-2), The left calf positioning coordinates are iC07-1 (XiC07-1, YiC07-1, ZiC07-1) and iC07-2 (XiC07-2, YiC07-2, ZiC07-2) and the right calf positioning coordinates are iC08-1 (XiC08-1, YiC08-1, ZiC08-1) and iC08-2 (XiC08-2, YiC08-2, ZiC08-2).

[0070] Where i is the unique number of the athlete; where A01, A02, ..., C08 are the numbers of the body structures; for the positioning coordinates of the limb parts, the coordinates of the two endpoints of the line segment are taken.

[0071] For example, C01-1 and C01-2 are the numbers of the two endpoints of the line segment corresponding to the left upper arm, and the rest are the same.

[0072] Step 3: Coordinate transformation: Establish a personal dynamic coordinate system, specifically: See also Figure 6 As shown, for each athlete's motion diagram, locate the polygon corresponding to the upper torso, obtain the geometric center coordinates iB02 (XiB02, YiB02, ZiB02) of the polygon, and record the plane where the polygon is located as the personal dynamic reference plane.

[0073] Draw a straight line in space, and make the straight line meet the following conditions: Condition 1, perpendicular to the personal dynamic reference plane; Condition 2, the straight line passes through the geometric center of the polygon; The straight line is recorded as reference normal vector one.

[0074] Furthermore, in the polygon corresponding to the upper torso, obtain the corner point closest to the head positioning coordinate iB01 (XiB01, YiB01, ZiB01), which is recorded as the feature point. Obtain the coordinates D (Xia, Yib, Zic) of the feature point. Draw a spatial straight line that satisfies the following conditions: Condition 1: the straight line is located in the personal dynamic reference plane; Condition 2: the straight line passes through the geometric center coordinates of the polygon; Condition 3: the straight line passes through the feature point.

[0075] The straight line is recorded as reference normal vector two.

[0076] A straight line perpendicular to the reference normal vector 2 is drawn in the personal dynamic reference plane, and is named reference normal vector 3. The reference normal vectors 1, 2 and 3 are perpendicular to each other.

[0077] With the centroid coordinates of each athlete as the coordinate base point, the reference normal vector 1 as the X-axis, the reference normal vector 2 as the Y-axis, and the reference normal vector 3 as the Z-axis, a spatial rectangular coordinate system is established, which is recorded as the personal dynamic coordinates of athlete i.

[0078] Where i is the unique number of the athlete, and t is the current moment. In the personal dynamic reference plane.

[0079] Furthermore, a coordinate transformation operation relationship is established between the perception coordinate system and the personal dynamic coordinate system, specifically: Get the coordinates of the geometric centroid of each athlete (Xi, Yi, Zi), and let the translation transformation matrix be: , get the angle θ1 between the reference normal vector 1 and the X axis in the perception coordinate system, get the angle θ2 between the reference normal vector 2 and the Y axis in the perception coordinate system, get the angle θ3 between the reference normal vector 3 and the Z axis in the perception coordinate system. Define the rotation transformation matrix as: .

[0080] Let the composite transformation matrix M be Given a point P (x, y, z) in the perception coordinate system, through the composite transformation matrix M, the new coordinates P ´ (x ´, y ´, z ´) of the point in the personal dynamic coordinate system can be calculated as follows: .

[0081] Furthermore, for each athlete, the collected positioning coordinates of each body structure of the athlete are transformed into the athlete's personal dynamic coordinate system in real time through the coordinate transformation operation relationship to obtain the relative coordinates of each body structure, including: The relative coordinates of the connection parts: the relative coordinates of the neck joint iA01' (XiA01', YiA01', ZiA01'), the relative coordinates of the left shoulder joint iA02' (XiA02', YiA02', ZiA02'), the relative coordinates of the right shoulder joint iA03' (XiA03', YiA03', ZiA03'), the relative coordinates of the waist joint iA04' (XiA04', YiA04', ZiA04'), the relative coordinates of the left hip joint iA05' (XiA05', YiA05', ZiA05'), the relative coordinates of the right hip joint iA06' (XiA06', YiA06', ZiA06'), the relative coordinates of the left elbow joint iA07' (XiA07', YiA07', ZiA07'). 07'), the left wrist joint's own relative coordinates iA08' (XiA08', YiA08', ZiA08'), the right elbow joint's own relative coordinates iA09' (XiA09', YiA09', ZiA09'), the right wrist joint's own relative coordinates iA10' (XiA10', YiA10', ZiA10'), the left knee joint's own relative coordinates iA11' (XiA11', YiA11', ZiA11'), the right knee joint's own relative coordinates iA12' (XiA12', YiA12', ZiA12'), the left ankle joint's own relative coordinates iA13' (XiA13', YiA13', ZiA13') and the right ankle joint's own relative coordinates iA14' (XiA14', YiA14', ZiA14'); The relative coordinates of the backbone part: the center coordinates of each polygon are the relative coordinates of the backbone part, including: the relative coordinates of the head iB01' (XiB01', YiB01', ZiB01'), the relative coordinates of the upper trunk iB02' (XiB02', YiB02', ZiB02'), the relative coordinates of the lower trunk (XiB03', YiB03', ZiB03'), the relative coordinates of the left palm iB04' (XiB04', YiB04', ZiB04'), the relative coordinates of the right palm iB05' (XiB05', YiB05', ZiB05'), the relative coordinates of the left sole iB06' (XiB06', YiB06', ZiB06') and the relative coordinates of the right sole iB07' (XiB07', YiB07', ZiB07'); The relative coordinates of the limbs: The coordinates of the two endpoints of each line segment are the relative coordinates of the limbs. Including: the relative coordinates of the left upper arm iC01-1' (XiC01-1', YiC01-1', ZiC01-1') and iC01-2' (XiC01-2', YiC01-2', ZiC01-2'), the relative coordinates of the right upper arm iC02-1' (XiC02-1', YiC02-1', ZiC02-1') and iC02-2' (XiC02-2', YiC02-2', ZiC02-2') ), the left forearm's own relative coordinates iC03-1' (XiC03-1', YiC03-1', ZiC03-1') and iC03-2' (XiC03-2', YiC03-2', ZiC03-2'), the right forearm's own relative coordinates iC04-1' (XiC04-1', YiC04-1', ZiC04-1') and iC04-2' (XiC04-2', YiC04-2', ZiC04-2') , the relative coordinates of the left thigh iC05-1' (XiC05-1', YiC05-1', ZiC05-1') and iC05-2' (XiC05-2', YiC05-2', ZiC05-2'), the relative coordinates of the right thigh iC06-1' (XiC06-1', YiC06-1', ZiC06-1') and iC06-2' (XiC06-2', YiC06-2', ZiC06-2'), The left calf's own relative coordinates are iC07-1' (XiC07-1', YiC07-1', ZiC07-1') and iC07-2' (XiC07-2', YiC07-2', ZiC07-2') and the right calf's own relative coordinates are iC08-1' (XiC08-1', YiC08-1', ZiC08-1') and iC08-2' (XiC08-2', YiC08-2', ZiC08-2').

[0082] It should be noted that if the perception coordinate system is used as the only reference system, when athletes are exercising, the displacement data of their own geometric center and the displacement data of various body parts will be coupled, affecting the accuracy of the data. The significance of coordinate transformation is to change the motion reference system of each athlete, accurately reflect the position and motion trajectory of each part of the athlete's body structure relative to its own geometric center, decompose the athlete's motion into the particle motion of its own geometric center and the relative motion of each body structure relative to the geometric center, thereby simplifying the complexity of the calculation.

[0083] For example, for a track and field athlete who is running a long-distance event, if data is collected in a perceptual coordinate system, only the relative movement speed of the athlete's various body parts on the track can be observed. However, this cannot accurately reflect the movement patterns of the athlete's legs, hips and other body parts during the competition, especially in different stages such as acceleration, maintaining speed and deceleration. In fact, when the athlete starts, the force of the leg pushing the ground is very large and the speed is very fast, but in the perceptual coordinate system, this high speed is only reflected in the overall forward acceleration, and it is impossible to specifically quantify the speed of each leg push.

[0084] Step 4: Numerical analysis For each athlete i, all body structures are uniquely numbered, and all connecting parts are uniquely numbered.

[0085] According to the relative coordinates of each body structure, the specific parts of all body structures are uniquely numbered, and the unique number is j1, j1=1, 2, ..., m1; wherein m is the total number of body parts.

[0086] Each connection part in the body structure is uniquely numbered, and the unique number is j2, j2=1, 2, ..., m2; wherein m2 is the total number of connection parts.

[0087] The relative coordinates of each body structure of the athlete collected are numerically analyzed from five perspectives: linear velocity, angular velocity, acceleration, amplitude and frequency of movement. Specifically: Motion linear velocity analysis: For each preset time interval t, the relative coordinates of each body structure are recorded. The relative coordinates of the current time E1 = (XE1, YE1, ZE1) are recorded, and the relative coordinates of the previous preset time interval t corresponding to the same body structure E2 = (XE2, YE2, ZE2) are obtained. By formula Calculate the linear displacement S(i1, j, T) of athlete i's body structure j at time T. Calculate the linear velocity VS(i, j1, T) of athlete i's body structure j1 at time T by the formula VS=S(i, j, T) / t; Analysis of angular velocity of motion: Obtain the self-relative coordinates of the connected parts. For each connected part, obtain the self-relative coordinates of the limb part and the backbone part connected to it in the motion diagram to form a connecting polyline. For the two self-relative coordinates of the end points of the limb part line segment, select the end point of the line segment according to the principle of the distal end point, and form a connecting polyline with the connected part.

[0088] It should be noted that the distal endpoint principle refers to taking the one that is farther away from the connecting part among the two line segment endpoints of the limb part.

[0089] For example, see Figure 7 As shown, for the left elbow joint, obtain its own relative coordinates iA07' (XiA07', YiA07', ZiA07'). In the motion diagram, obtain the coordinates of the limb parts connected to it, that is, the left upper arm's own relative coordinates iC01-1' (XiC01-1', YiC01-1', ZiC01-1') and iC01-2' (XiC01-2', YiC01-2', ZiC01-2') and the left forearm's own relative coordinates iC03-1' (XiC03-1', YiC03-1', ZiC03-1') and iC03-2' (XiC03-2', YiC03-2', ZiC03-2'). Form a connecting line left upper arm iC01-1'-left elbow joint iA07'-left forearm iC03-2'. The distance between point iC03-1' and iA07' is greater than the distance between point iC03-2' and iA07'; the distance between point iC03-2' and iA07' is greater than that between point iC03-1'.

[0090] At every preset time interval t, the angles of each connecting polyline are obtained through spatial coordinate calculation. The specific process is as follows: In the connecting polyline, let the two end points be F1 and F2, and let the middle connecting point be F0.

[0091] Get their coordinates F0 (XF0, YF0, ZF0), F1 (XF1, YF1, ZF1) and F3 (XF3, YF3, ZF3) respectively. Calculate the connection line angle θ(i, j2, T) corresponding to the connection part j2 of athlete i at the current moment, where i is the unique number of each athlete; j2 is the unique number of each connection part; and T is the current moment. Get the previous preset time interval t and calculate the connection line angle θ(i, j2, Tt) corresponding to the connection part j2 of athlete i.

[0092] The changing speed of the angle θ(i, j2, T) of the connecting broken line is calculated by the formula vθ(i, j2, T)=θ[(i, j2, T)-θ(i, j2, Tt)] / t, which is recorded as the angular velocity of the angle of each connecting part.

[0093] Motion acceleration analysis: Based on the calculation results of the motion linear velocity analysis, the motion linear velocity VS (i, j1, T) of the athlete i's body structure j1 at time T and the motion linear velocity VS´ (i, j1, Tt) at the previous preset time interval t are obtained, and the linear acceleration of the athlete i's body structure j1 at time T is calculated by the formula aS (i, j2, T) = [VS (i, j1, T) - VS´ (i, j1, Tt)] / t.

[0094] Based on the calculation results of the motion angular velocity analysis, the motion angular velocity Vθ(i, j2, T) of the connecting part j2 of athlete i at time T and the motion linear velocity Vθ´(i, j2, Tt) at the previous preset time interval t are obtained, and the linear acceleration of the connecting part j2 of athlete i at time T is calculated by the formula aθ(i, j2, T)=[Vθ(i, j2, T)-Vθ´(i, j2, Tt)] / t.

[0095] Motion amplitude analysis: preset an amplitude monitoring period T1=n1×t, where n1 is a preset period base, n1 is a positive integer greater than 0; and t is a preset time interval.

[0096] Based on the analysis and calculation results of the linear velocity of motion, every amplitude monitoring period T1, the linear displacement of the body structure j of athlete i at each moment is obtained and its maximum value Smax(i1, j, T) is extracted, which is recorded as the motion amplitude of the body structure j of athlete i within the amplitude monitoring period T1.

[0097] Motion frequency analysis: Preset a frequency monitoring period T2=n2×t, where n2 is the preset frequency base, n2 is a positive integer greater than 0; and t is the preset time interval.

[0098] A minimum displacement benchmark Smin is preset. Based on the analysis and calculation results of the linear velocity of the movement, the linear displacement of the body structure j of the athlete i at each moment is obtained. Subsequently, the linear displacement greater than the minimum displacement benchmark Smin is marked as a valid linear displacement. Every amplitude monitoring period T1, the number of valid linear displacements Q is counted and recorded as the movement frequency.

[0099] The analysis and calculation of the linear velocity, angular velocity, acceleration, amplitude and frequency are repeatedly performed at preset time intervals t, so that the analysis and calculation process traverses the unique numbers i of all athletes and the unique numbers j1 of all body structures.

[0100] Step 5: Visualization of numerical values; The data visualization of the motion diagram is as follows: The specific values ​​of the linear velocity, angular velocity, linear acceleration, angular acceleration, amplitude and frequency of movement are marked in the movement diagram according to the unique number i of the athlete and the unique number j1 of all body structures.

[0101] A set of data thresholds are preset, including a motion linear velocity threshold, a motion angular velocity threshold, a linear acceleration threshold, an angular acceleration threshold, a motion amplitude threshold, and a motion frequency threshold. Specific values ​​exceeding their own thresholds are highlighted in the marked motion diagram.

[0102] Furthermore, the motion diagrams processed by data visualization are separated and saved at time intervals t, providing a playback window for the motion diagrams over time.

[0103] Step 6: Data preservation; The positioning coordinates of all body structures of each athlete, the positioning coordinates after coordinate conversion, the linear velocity, angular velocity, acceleration, amplitude and frequency of movement obtained through analysis and calculation are saved.

[0104] Furthermore, the motion diagram with specific numerical annotations and highlights is saved.

[0105] See also Figure 8 and Fig. 9 As shown, a motion data processing device based on computer vision comprises: A tripod 1 is provided with a circular support 2 at the upper end of the tripod, and an industrial-grade high-definition camera 3 for shooting is provided at the upper end of the support 2. A freely extendable support leg 1 101, a support leg 2 102 and a support leg 3 103 are provided at the lower end of the tripod 1. The support legs 101, 102 and 103 intersect at the upper end of the tripod 1 and are rigidly connected by a triangular horizontal limit plate 104, and a limit hole 105 is provided at the top end of the tripod 1.

[0106] The support platform 2 includes a coupling component 1 202 at the upper end and a circular support platform plane 201 at the lower end. The lower end of the support platform plane 201 is connected to the limiting hole 105 in the tripod 1 to limit the displacement of the support platform 2 and ensure that the support platform plane rotates with the limiting hole 105 as the center during use. The coupling component 1 202 forms a limiting connection between the support platform plane 201 and the camera pan 301 in the industrial-grade high-definition camera 3, limiting the displacement of the industrial-grade high-definition camera 3 and ensuring that the industrial-grade high-definition camera 3 rotates with the coupling component 1 202 as the center during use.

[0107] The industrial-grade high-definition camera 3 includes a camera platform 301 at the lower end. The camera platform 301 is connected to the camera body 302 by a coupling component 303 on the side of the camera body 302 to form a limiting connection.

[0108] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0109] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms "a", "an", and "the" are intended to include plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations; The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A motion data processing method based on computer vision, characterized in that: The following steps are involved: Step 1: Layout of the sensing network; arrange several cameras with fixed positions, and constrain the arrangement positions of each camera through numerical limit calculation; adjust each industrial-grade high-definition camera so that its sight direction faces the sports field; then, take the athlete's face image and retrieve the athlete's face recognition feature map from the database, perform face recognition comparison based on computer vision technology, match the specific athlete according to the face recognition comparison result and assign a unique number; Step 2: Image processing; joint recognition and human body model construction; video recording and frame-by-frame decomposition by various industrial-grade high-definition cameras to obtain the motion images of each athlete at each time t; distortion correction and image noise reduction processing are performed on the motion images; and feature extraction of the motion images is performed based on computer vision technology; Step 3: Coordinate transformation: Establish a personal dynamic coordinate system, and transform the positioning coordinates from the perception coordinate system to the personal dynamic coordinate system through coordinate transformation operation: Step 4: Numerical analysis: For each athlete, all body structures are uniquely numbered, and the relative coordinates of each body structure of the athlete collected are numerically analyzed from five perspectives: linear velocity, angular velocity, acceleration, amplitude, and frequency of movement. Specifically: Step 5: Visualize numerical values; perform data visualization on the motion diagram; Step 6: Data preservation; save the positioning coordinates of all body structures of each athlete, the positioning coordinates after coordinate conversion, the linear velocity, angular velocity, acceleration, amplitude and frequency of movement obtained through analysis and calculation; save the simple movement diagram with specific numerical annotations and highlights.

2. The motion data processing method based on computer vision according to claim 1, characterized in that: The numerical limit calculation of each camera is as follows: Set up industrial-grade high-definition camera 1 at preset position 1, set up industrial-grade high-definition camera 2 at preset position 2, and set up industrial-grade high-definition camera 3 at preset position 3; let preset position 1 be the coordinate origin O1, establish a three-dimensional rectangular coordinate system with due north as the Y axis, due east as the Y axis, and vertical direction as the Z axis, and record it as the perception coordinate system; obtain the coordinates O2 (Xo2, Yo2, Zo2) and O3 (Xo3, Yo3, Zo3) of preset positions such as preset position 2 and preset position 3 in the perception coordinate system; For preset positions 1, 2 and 3, preset distance thresholds dmax and dmin, and preset maximum angles θmax and θmin; Connect points O1, O2 and O3 in sequence to form a triangle; let the coordinates of point O1 be (Xo1, Yo1, Zo1) by the formula Calculate the distance a between the preset positions 1 and 2, the distance b between the preset positions 1 and 3, and the distance c between the preset positions 2 and 3; By distance constraint Limiting the distances a, b and c; By formula Calculate the angles θa, θb and θc of the triangle formed by the preset positions one, two and three; By Angle Constraint The angles θa, θb and θc are limited; When the number of preset positions exceeds three, that is, when the number of industrial-grade high-definition cameras installed is greater than three, adjacent preset position coordinate points are sequentially connected to form a plurality of triangles; and the above-mentioned numerical limit calculation is performed for each triangle.

3. The motion data processing method based on computer vision according to claim 1, characterized in that: The specific process of feature extraction for motion images is as follows: Based on computer vision technology, feature extraction is performed on the sports image to construct a three-dimensional digital twin model of each athlete, and each body structure of the athlete is located in the three-dimensional digital twin model through a key point detection neural network model; the body structure is divided into a connection part, a backbone part and a limb part according to the movement mode and numbered respectively: The three-dimensional digital twin model of each athlete is geometrically abstracted, and the specific form of the human body is abstracted into a combination of a series of geometric figures; the connecting parts are simplified by circles, the backbone parts are simplified by polygons, and the limbs are simplified by line segments; and then a motion diagram is generated; At every preset time interval t, characteristic coordinate points are extracted from the motion diagram, motion data is quantified, and the geometric centroid of the motion diagram of each athlete as a whole is located, that is, the geometric center of the image composed of all circles and polygons in each motion diagram; the coordinates (xi, yi, zi) of the geometric centroid of the motion diagram of each athlete in the perception coordinate system are recorded at every preset time interval t, and recorded as positioning coordinates; for the motion diagram of each athlete, the positioning coordinates of each body structure are respectively obtained at every preset time interval t in the perception coordinate system, specifically: For the connecting part, the coordinates of the center of each circle are used as the positioning coordinates of the connecting part; for the backbone part, the coordinates of the center of each polygon are used as the positioning coordinates of the backbone part; for the limb part, the coordinates of the two endpoints of each line segment are used as the positioning coordinates of the limb part.

4. The motion data processing method based on computer vision according to claim 1, characterized in that: The specific process of establishing a personal dynamic coordinate system is as follows: For each athlete's motion diagram, locate the polygon corresponding to the upper torso; obtain the geometric center coordinates iB02 (XiB02, YiB02, ZiB02) of the polygon; record the plane where the polygon is located as the personal dynamic reference plane; Draw a straight line in space, and make the straight line meet the following conditions: condition 1, perpendicular to the personal dynamic reference plane; condition 2, the straight line passes through the geometric center of the polygon; the straight line is recorded as the reference normal vector 1; Obtain the corner point closest to the head positioning coordinate iB01 (XiB01, YiB01, ZiB01) in the polygon corresponding to the upper torso, and record it as a feature point; obtain the coordinates D (Xia, Yib, Zic) of the feature point; draw a spatial straight line, and make the straight line meet the following conditions: Condition 1, the straight line is located in the personal dynamic reference plane; Condition 2, the straight line passes through the geometric center coordinates of the polygon; Condition 3, the straight line passes through the feature point; The straight line is recorded as reference normal vector 2; Draw a straight line perpendicular to the reference normal vector 2 in the personal dynamic reference plane, named as reference normal vector 3; the reference normal vectors 1, 2 and 3 are perpendicular to each other; With the centroid coordinates of each athlete as the coordinate base point, the reference normal vector 1 as the X-axis, the reference normal vector 2 as the Y-axis, and the reference normal vector 3 as the Z-axis, a spatial rectangular coordinate system is established, which is recorded as the personal dynamic coordinates of athlete i.

5. The method for processing motion data based on computer vision according to claim 1, characterized in that: The coordinate transformation operation from the perception coordinate system to the personal dynamic coordinate system is as follows: Get the coordinates of the geometric centroid of each athlete (Xi, Yi, Zi), and let the translation transformation matrix be: , get the angle θ1 between the reference normal vector 1 and the X axis in the perception coordinate system, get the angle θ2 between the reference normal vector 2 and the Y axis in the perception coordinate system, get the angle θ3 between the reference normal vector 3 and the Z axis in the perception coordinate system. Define the rotation transformation matrix as: ； Let the composite transformation matrix M be ; Given a point P (x, y, z) in the perception coordinate system, the new coordinates P ´ (x ´, y ´, z ´) of the point in the personal dynamic coordinate system can be calculated by the composite transformation matrix M as follows: ; For each athlete, the collected positioning coordinates of each body structure of the athlete are transformed into the athlete's personal dynamic coordinate system in real time through the coordinate transformation operation relationship, and the relative coordinates of each body structure are obtained, including: The relative coordinates of the connecting part, the relative coordinates of the backbone part and the relative coordinates of the limb part.

6. The method for processing motion data based on computer vision according to claim 1, characterized in that: The specific process of numerical analysis from the five perspectives of linear velocity, angular velocity, acceleration, amplitude and frequency of motion is as follows: Motion linear velocity analysis: For each preset time interval t, the relative coordinates of each body structure are recorded; the relative coordinates of the current time E1 = (XE1, YE1, ZE1) are recorded, and the relative coordinates of the same body structure at the previous preset time interval t are obtained. Calculate the linear displacement S(i1, j, T) of the body structure j of athlete i at time T; calculate the linear velocity VS(i, j1, T) of the body structure j1 of athlete i at time T by the formula VS=S(i, j, T) / t; Analysis of angular velocity of motion: obtain the relative coordinates of the connected part; for each connected part, obtain the relative coordinates of the limb part and the backbone part connected to it in the motion diagram to form a connecting polyline; for the two relative coordinates of the endpoints of the limb part line segment, select the endpoints of the line segment according to the principle of the distal endpoint, and form a connecting polyline with the connected part; At every preset time interval t, the angles of each connecting polyline are obtained through spatial coordinate calculation. The specific process is as follows: In the connecting polyline, let the two endpoints be F1 and F2, and let the middle connecting point be F0; Get their coordinates F0 (XF0, YF0, ZF0), F1 (XF1, YF1, ZF1) and F3 (XF3, YF3, ZF3) respectively; through the formula Calculate the connection broken line angle θ(i, j2, T) corresponding to the connection part j2 of the athlete i at the current moment, where i is the unique number of each athlete; j2 is the unique number of each connection part; where T is the current moment; obtain the previous preset time interval t, and calculate the connection broken line angle θ(i, j2, Tt) corresponding to the connection part j2 of the athlete i; The change rate of the angle θ(i, j2, T) of the connecting line is calculated by the formula vθ(i, j2, T) = θ[(i, j2, T) - θ(i, j2, Tt)] / t, which is recorded as the angular velocity of the angle of each connecting part; Motion acceleration analysis: Based on the calculation result of the motion linear velocity analysis, the motion linear velocity VS (i, j1, T) of the body structure j1 of the athlete i at the time T and the motion linear velocity VS´ (i, j1, Tt) at the last preset time interval t are obtained, and the linear acceleration of the body structure j1 of the athlete i at the time T is calculated by the formula aS (i, j2, T) = [VS (i, j1, T) - VS´ (i, j1, Tt)] / t; Based on the calculation result of the motion angular velocity analysis, the motion angular velocity Vθ(i, j2, T) of the connection part j2 of the athlete i at the time T and the motion linear velocity Vθ´(i, j2, Tt) at the last preset time interval t are obtained, and the linear acceleration of the connection part j2 of the athlete i at the time T is calculated by the formula aθ(i, j2, T)=[Vθ(i, j2, T)-Vθ´(i, j2, Tt)] / t; Motion amplitude analysis: preset an amplitude monitoring period T1=n1×t, where n1 is the preset period base, and n1 is a positive integer greater than 0; Where t is the preset time interval; Based on the analysis and calculation results of the motion linear velocity, every amplitude monitoring period T1, the linear displacement of the body structure j of athlete i at each moment is obtained and its maximum value Smax(i1, j, T) is extracted, which is recorded as the motion amplitude of the body structure j of athlete i in the amplitude monitoring period T1; Motion frequency analysis: preset a frequency monitoring period T2=n2×t, where n2 is the preset frequency base, and n2 is a positive integer greater than 0; Where t is the preset time interval; A minimum displacement benchmark Smin is preset; based on the analysis and calculation results of the linear velocity of the movement, the linear displacement of the body structure j of the athlete i at each moment is obtained; Subsequently, the line displacements that are greater than the minimum displacement reference Smin are marked as valid line displacements; every amplitude monitoring period T1, the number of valid line displacements Q is counted and recorded as the movement frequency; The analysis and calculation of the linear velocity, angular velocity, acceleration, amplitude and frequency are repeatedly performed at preset time intervals t, so that the analysis and calculation process traverses the unique numbers i of all athletes and the unique numbers j1 of all body structures.

7. The motion data processing method based on computer vision according to claim 1, characterized in that: The specific process of data visualization of the motion diagram is as follows: The specific values ​​of the linear velocity, angular velocity, linear acceleration, angular acceleration, amplitude and frequency of movement are marked in the movement diagram according to the unique number i of the athlete and the unique number j1 of all body structures; A set of data thresholds are preset, including a motion linear velocity threshold, a motion angular velocity threshold, a linear acceleration threshold, an angular acceleration threshold, a motion amplitude threshold, and a motion frequency threshold; specific values ​​exceeding their own thresholds are highlighted in the marked motion diagram; The motion diagram that has been processed by data visualization is separated and saved at time intervals t, providing a playback window for the motion diagram interpretation over time.

8. A motion data processing device based on computer vision, characterized in that: The method for processing motion data based on computer vision according to any one of claims 1 to 7 comprises: The camera module takes the athlete's facial image and retrieves the athlete's facial recognition feature map from the database, performs face recognition comparison based on computer vision technology, matches the specific athlete according to the face recognition comparison result and assigns a unique number; uses various industrial-grade high-definition cameras to record and photograph and decompose frame by frame to obtain the motion image of each athlete at each time t; performs distortion correction and image noise reduction processing on the motion image; and extracts features from the motion image based on computer vision technology.

9. The motion data processing device based on computer vision according to claim 8, characterized in that: It includes a tripod (1), a support platform (2) and an industrial-grade high-definition camera (3); An industrial-grade high-definition camera (3) for shooting is installed at the upper end of the support platform (2); a first support leg (101), a second support leg (102) and a third support leg (103) are provided at the lower end of the tripod (1); the first support leg (101), the second support leg (102) and the third support leg (103) intersect at the upper end of the tripod (1) and are rigidly connected via a triangular horizontal limiting plate (104); a limiting hole (105) is provided at the top end of the tripod (1); The support platform (2) comprises a coupling component 1 (202) at the upper end and a circular support platform plane (201) at the lower end; the lower end of the support platform plane (201) is connected to a limiting hole (105) in the tripod (1) to limit the displacement of the support platform (2) and ensure that the support platform plane rotates with the limiting hole (105) as the center during use; the coupling component 1 (202) forms a limiting connection between the support platform plane (201) and a camera pan / tilt (301) in the industrial-grade high-definition camera (3) to limit the displacement of the industrial-grade high-definition camera (3) and ensure that the industrial-grade high-definition camera (3) rotates with the coupling component 1 (202) as the center during use; The industrial-grade high-definition camera (3) comprises a camera pan platform (301) at the lower end; through a second coupling component (303) on the side of the camera body (302), the camera pan platform (301) is connected to the camera body (302) in a limited position.

Images