Linear Rigid Body Three-Dimensional Space Motion Tracking System and Method Based on Edge Computing

Abstract

This invention discloses a system and method for tracking the three-dimensional motion of linear rigid bodies based on edge computing. The method comprises: a detection phase, acquiring feature information containing the number of linear rigid bodies and their rotational fulcrums from video stream data; and a tracking phase, utilizing this feature information to detect linear rigid bodies in specific areas and obtain their rotation angles in three-dimensional space, thereby completing the motion tracking of the linear rigid bodies in three-dimensional space. The invention utilizes computer vision technology and requires only the addition of a monocular camera and an inertial sensor, enabling real-time motion tracking of linear rigid bodies in three dimensions in conjunction with an edge processor.

Description

Technical Field

[0001] The present invention belongs to the field of edge computing and visual perception, and specifically relates to a linear rigid body three-dimensional space motion tracking system and method based on edge computing. Background Art

[0002] In today's era of rapid IoT development, with the continuous improvement of edge node computing power, more and more edge devices are emerging. These edge devices support basic computing, and some can even use lightweight deep neural network reasoning. Leveraging the computing resources of these edge nodes, data can be processed locally, achieving faster responses, avoiding unnecessary data transmission, and ensuring data privacy.

[0003] In factories and other production environments, linear rigid structures such as knife switches and robotic arms are very common. The proper operation of these facilities can significantly impact factory efficiency and safety. This is especially true when designing delicate operations, which require real-time detection of the linear rigid body's rotation angle in three dimensions. In current factory systems, the detection of these linear rigid bodies can generally be categorized into three approaches: manual inspection, sensor-based methods, and computer vision-based methods.

[0004] The manual inspection method is simple and direct, but for some factories with remote locations or linear rigid bodies installed in places that are inconvenient for staff to check, the cost of manual inspection is too high. In addition, manual inspection can only judge whether the rotation of the linear rigid body is in place based on the staff's experience, and the accuracy cannot be guaranteed.

[0005] While sensor-based methods can accurately detect the motion of linear rigid bodies in some scenarios, they are mostly contact-based, limiting their use cases. For example, detecting knife switches in high-voltage transmission lines poses safety risks. Furthermore, sensors are limited in the dimensions they can detect, making it difficult to track the motion of linear rigid bodies in three-dimensional space.

[0006] Computer vision-based methods fall into two main categories: traditional computer vision-based methods and deep neural network-based methods. Edge device processors with limited computing power struggle to implement real-time, multi-target, three-dimensional motion tracking using deep neural networks. Traditional computer vision-based methods, unable to obtain depth information from two-dimensional images, also struggle to accurately calculate the rotation angle of linear rigid bodies in three dimensions.

[0007] Therefore, based on the above considerations, it is necessary to propose a linear rigid body three-dimensional space motion tracking system based on edge computing, which combines the video stream data of linear rigid body motion obtained by the camera and the camera posture information obtained by the inertial sensor to realize real-time multi-target linear rigid body three-dimensional space motion tracking on edge devices with limited computing power. Summary of the Invention

[0008] To address the shortcomings of the aforementioned prior art, the present invention aims to provide a system and method for tracking the three-dimensional motion of linear rigid bodies based on edge computing, addressing the existing inability to use edge processors to track the motion of linear rigid bodies in real time. This system utilizes computer vision technology and, by simply adding a monocular camera and an inertial sensor, can work with edge processors to achieve real-time tracking of the motion of linear rigid bodies in three dimensions.

[0009] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0010] The present invention provides a linear rigid body three-dimensional space motion tracking system based on edge computing, comprising: a remote detection all-in-one machine, a server platform and a client;

[0011] The remote detection all-in-one machine is used to obtain real-time linear rigid body motion video stream data, track the linear rigid body's motion in three-dimensional space in the video stream data, obtain the linear rigid body's rotation angle in three-dimensional space, and send the result data to the server platform;

[0012] The server platform is used to obtain the result data sent by the remote detection machine in real time and store it;

[0013] The client is used to obtain result data from the server platform in real time and query historical result data.

[0014] Furthermore, the remote detection integrated machine includes: a camera, an inertial sensor, a processor and a communication interface;

[0015] Camera, used to collect linear rigid body motion video stream data;

[0016] Inertial sensor, used to obtain the angle between the camera plane and the linear rigid body in the reference frame;

[0017] A processor is used to obtain the rotation angle of the linear rigid body in three-dimensional space based on the video stream data collected by the camera, and complete its motion tracking in three-dimensional space;

[0018] Communication interface, used to send the rotation angle results of the linear rigid body in three-dimensional space to the server platform.

[0019] The present invention provides a linear rigid body three-dimensional space motion tracking method based on edge computing, based on the above system, including: a detection stage and a tracking stage; wherein,

[0020] In the detection phase, feature information including the number of linear rigid bodies and rotation pivot points is obtained from the video stream data;

[0021] In the tracking stage, the linear rigid body is detected in a specific area using its characteristic information, and its rotation angle in three-dimensional space is obtained to complete the motion tracking of the linear rigid body in three-dimensional space.

[0022] Furthermore, the specific steps of the detection stage are:

[0023] (1) Receive linear rigid body video stream data collected by a camera, and extract the first frame F1 and the second frame F2 from the video stream data;

[0024] (2) Using the frame difference method, the motion region of the linear rigid body is extracted from the two adjacent frames F1 and F2 to form a mask M;

[0025] (3) Extract all feature points from the mask M and divide the feature points to obtain the number of linear rigid bodies N' and a set of feature point sets P corresponding to each linear rigid body i i 0 , i∈[1,N']; using linear fitting, we get a distance feature point set P i 0 The shortest straight line between all feature points

[0026] (4) For each of the subsequent N frames, use the optical flow method to obtain P i 0 The feature point set P matched in the subsequent n-th frame i n , and use linear fitting to get a distance feature point set P i 0 The shortest straight line between all feature points

[0027] (5) For each linear rigid body, combine N straight lines Fit the nearest intersection point of N straight lines as the rotation fulcrum K of the linear rigid body i i .

[0028] Furthermore, the feature point division method in step (3) is specifically as follows:

[0029] (31) Initially, all feature points are in the undivided point set I;

[0030] (32) Each time a feature point p is taken from the unpartitioned point set I, and a feature point set P is found i , so that P i The center point c i The closest distance to the feature point p, and the distance is less than / equal to the threshold T, the feature point p is put into the feature point set P i If there is no feature point set P that meets the conditions i , then create a new feature point set P n+1 , n is the number of existing feature point sets, and the feature point p is put into the feature point set P n+1 ; Use the Euclidean distance center value of all feature points in the feature point set as the center point c of the feature point set i ;

[0031] (33) When the undivided point set I is empty, the number of feature point sets obtained is the number of linear rigid bodies N'. Each feature point set represents a feature point set P corresponding to a linear rigid body i. i .

[0032] Furthermore, the specific steps of the tracking stage are:

[0033] (6) For the input linear rigid body video stream, use the rotation fulcrum K i The line detection method with a known endpoint detects the rotation angle RP of the linear rigid body i in the camera plane direction i and length S i ;

[0034] (7) The first frame of the tracking phase is used as the reference frame T of the linear rigid body i i , and using the reference frame T i Calculate the base length of the linear rigid body i as S' i ;

[0035] (8) The length S of the current frame of the linear rigid body i i and its reference length S′ i For comparison, obtain the rotation angle RS of the linear rigid body in the direction perpendicular to the camera plane i ;

[0036] (9) Combined with the rotation angle RP of the linear rigid body in the direction of the camera plane i and the rotation angle RS of the linear rigid body in the direction perpendicular to the camera plane i , complete the motion tracking of the linear rigid body i in three-dimensional space.

[0037] Furthermore, the linear rigid body detection method in step (6) is specifically as follows:

[0038] (61) Rotate the fulcrum K iAs an endpoint of the linear rigid body i;

[0039] (62) According to the rotation angle RP′ of the linear rigid body in the camera plane direction in the previous frame and the angle threshold r, i As the starting point, a straight line detection with an accuracy of s is performed within the angle range of [RP′-r, RP′+r] to obtain the linear rigid body i;

[0040] (63) The mask M of the linear rigid body i is obtained by using the difference between the current frame and the previous frame. i , linear rigid body i and mask M i Do the intersection to get the detection results of the linear rigid body i, including the rotation angle RP of the linear rigid body i in the camera plane i and length S i .

[0041] Furthermore, the step (62) specifically includes:

[0042] (621) According to the two frames F before the current frame F0 -2 and F -1 The rotation angles RP″ and RP′ of the linear rigid body in the camera plane direction are used to obtain the current motion velocity V of the linear rigid body i:

[0043] V = RP′-RP″;

[0044] (622) According to the motion speed V, the detection angle range is optimized to [RP′-min(αV,-σ), RP′+max(αV,σ)], where α is a hyperparameter used to cope with the change of the linear rigid body's rotation speed; σ is the detection angle accuracy unit used to timely detect the change of the linear rigid body's motion direction;

[0045] (623) The previous frame F of the current frame -1 The running time t of the system is compared with the single-frame running time t′ required by the real-time performance of the system, and the accuracy s of the straight line detection in the straight line detection is dynamically adjusted; if t>t′, the accuracy s is reduced; if t≤t′, s remains unchanged;

[0046] (624) From the rotation fulcrum K i At the beginning, along the angle θ∈[RP′-min(αV,-σ), RP′+max(αV,σ)], every s pixels are judged according to the gray value of the pixel to see whether it is a continuous straight line, and the length of the straight line at the angle θ is recorded S θ ;

[0047] (625) Repeat step (624) until all angles θ in the angle range [RP′-min(αV,-σ),RP′+max(αV,σ)] are detected, and the longest straight line length S is taken. θ′and its corresponding angle θ′, with the rotation fulcrum K i Initially, the angle is θ′ and the length is S θ′ As the detection result of linear rigid body i.

[0048] Furthermore, the method for calculating the reference length in step (7) is specifically as follows:

[0049] According to step (6), the length S of the linear rigid body i in the camera plane in the reference frame is obtained i , use the inertial sensor to obtain the angle k between the linear rigid body i and the camera plane in the reference frame i , obtain the reference length S′ i :

[0050]

[0051] Furthermore, the calculation method in step (8) is specifically as follows:

[0052] The reference frame T i The deflection angle perpendicular to the camera plane is considered as RS′ i =0; when the length of the linear rigid body i is S i When the deflection angle RS in the direction perpendicular to the camera plane is obtained i :

[0053]

[0054] Beneficial effects of the present invention:

[0055] 1. High-precision real-time monitoring: High-precision real-time monitoring of linear rigid bodies, about 30 frames per second can meet the real-time motion tracking of linear rigid bodies in three-dimensional space;

[0056] 2. Reduced computational overhead: By leveraging the invariant position of the rotational pivot point of a linear rigid body, a combined detection-tracking approach, and an algorithm optimization method that introduces adaptive adjustment of thresholds and detection granularity during the tracking phase, the real-time performance of the algorithm is improved without compromising detection accuracy.

[0057] 3. Strong robustness: Using the frame difference method as a mask to distinguish linear rigid bodies from the background can effectively deal with occlusion, and using the feature point set partitioning method, it can handle any number of prior rigid body detections;

[0058] 4. Low cost: Based on the edge processor, you only need to add a cheap monocular camera and an inertial sensor to run linear rigid body detection and three-dimensional motion tracking, which is very low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1This is a system architecture diagram of the present invention;

[0060] Figure 2 is a flow chart of the method of the present invention;

[0061] Figure 3 This is a schematic diagram of the rotation fulcrum detection;

[0062] Figure 4 Schematic diagram for algorithm optimization;

[0063] Figure 5 Schematic diagram for calculating the benchmark length of the reference frame;

[0064] Figure 6 Schematic diagram for vertical angle calculation. DETAILED DESCRIPTION

[0065] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and drawings. The contents mentioned in the embodiments are not intended to limit the present invention.

[0066] Reference Figure 1 As shown, a linear rigid body three-dimensional space motion tracking system based on edge computing of the present invention includes: a remote detection all-in-one machine, a server platform and a client;

[0067] The remote detection all-in-one machine is used to obtain real-time linear rigid body motion video stream data, track the linear rigid body's motion in three-dimensional space in the video stream data, obtain the linear rigid body's rotation angle in three-dimensional space, and send the result data to the server platform;

[0068] The server platform is used to obtain the result data sent by the remote detection machine in real time and store it;

[0069] The client is used to obtain result data from the server platform in real time and query historical result data.

[0070] Specifically, the remote detection integrated machine includes: a camera, an inertial sensor, a processor and a communication interface;

[0071] Camera, used to collect linear rigid body motion video stream data;

[0072] Inertial sensor, used to obtain the angle between the camera plane and the linear rigid body in the reference frame;

[0073] A processor is used to obtain the rotation angle of the linear rigid body in three-dimensional space based on the video stream data collected by the camera, and complete its motion tracking in three-dimensional space;

[0074] Communication interface, used to send the rotation angle results of the linear rigid body in three-dimensional space to the server platform.

[0075] Reference Figure 2 As shown, a linear rigid body three-dimensional space motion tracking method based on edge computing of the present invention is based on the above system and includes: a detection stage and a tracking stage; wherein,

[0076] In the detection phase, feature information including the number of linear rigid bodies and rotation pivot points is obtained from the video stream data;

[0077] In the tracking stage, the linear rigid body is detected in a specific area using its characteristic information, and its rotation angle in three-dimensional space is obtained to complete the motion tracking of the linear rigid body in three-dimensional space.

[0078] Specifically, the specific steps of the detection stage are:

[0079] (1) Receive linear rigid body video stream data collected by a camera, and extract the first frame F1 and the second frame F2 from the video stream data;

[0080] (2) Using the frame difference method, the motion region of the linear rigid body is extracted from the two adjacent frames F1 and F2 to form a mask M;

[0081] (3) Extract all feature points from the mask M and divide the feature points to obtain the number of linear rigid bodies N' and a set of feature point sets P corresponding to each linear rigid body i i 0 , i∈[1,N']; using linear fitting, we get a distance feature point set P i 0 The shortest straight line between all feature points

[0082] (4) For each of the subsequent N frames, use the optical flow method to obtain P i 0 The feature point set P matched in the subsequent n-th frame i n , and use linear fitting to get a distance feature point set P i 0 The shortest straight line between all feature points

[0083] (5) For each linear rigid body, combine N straight lines Fit the nearest intersection point of N straight lines as the rotation fulcrum K of the linear rigid body i i , refer to Figure 3 shown.

[0084] The feature point division method in step (3) is specifically as follows:

[0085] (31) Initially, all feature points are in the undivided point set I;

[0086] (32) Each time a feature point p is taken from the unpartitioned point set I, and a feature point set P is found i , so that P i The center point c i The closest distance to the feature point p, and the distance is less than / equal to the threshold T, the feature point p is put into the feature point set P i If there is no feature point set P that meets the conditions i , then create a new feature point set P n+1 , n is the number of existing feature point sets, and the feature point p is put into the feature point set P n+1 ; Use the Euclidean distance center value of all feature points in the feature point set as the center point c of the feature point set i ;

[0087] (33) When the undivided point set I is empty, the number of feature point sets obtained is the number of linear rigid bodies N'. Each feature point set represents a feature point set P corresponding to a linear rigid body i. i .

[0088] Specifically, the specific steps of the tracking stage are:

[0089] (6) For the input linear rigid body video stream, use the rotation fulcrum K i The line detection method with a known endpoint detects the rotation angle RP of the linear rigid body i in the camera plane direction i and length S i ;

[0090] (7) The first frame of the tracking phase is used as the reference frame T of the linear rigid body i i , and using the reference frame T i Calculate the base length of the linear rigid body i as S' i ;

[0091] (8) The length S of the current frame of the linear rigid body i i and its reference length S′ i For comparison, obtain the rotation angle RS of the linear rigid body in the direction perpendicular to the camera plane i ;

[0092] (9) Combined with the rotation angle RP of the linear rigid body in the direction of the camera plane i and the rotation angle RS of the linear rigid body in the direction perpendicular to the camera plane i , complete the motion tracking of the linear rigid body i in three-dimensional space.

[0093] The linear rigid body detection method in step (6) is specifically as follows:

[0094] (61) Rotate the fulcrum Ki As an endpoint of the linear rigid body i;

[0095] (62) According to the rotation angle RP′ of the linear rigid body in the camera plane direction in the previous frame and the angle threshold r, i As the starting point, a straight line detection with an accuracy of s is performed within the angle range of [RP′-r, RP′+r] to obtain the linear rigid body i;

[0096] (63) The mask M of the linear rigid body i is obtained by using the difference between the current frame and the previous frame. i , linear rigid body i and mask M i Do the intersection to get the detection results of the linear rigid body i, including the rotation angle RP of the linear rigid body i in the camera plane i and length S i .

[0097] Wherein, the step (62) specifically includes:

[0098] (621) According to the two frames F before the current frame F0 -2 and F -1 The rotation angles RP″ and RP′ of the linear rigid body in the camera plane direction are used to obtain the current motion velocity V of the linear rigid body i:

[0099] V = RP′-RP″;

[0100] (622) According to the motion speed V, the detection angle range is optimized to [RP′-min(αV,-σ), RP′+max(αV,σ)], where α is a hyperparameter used to cope with the transformation of the linear rigid body's rotation speed; σ is the unit of detection angle accuracy used to timely detect the transformation of the linear rigid body's motion direction, refer to Figure 4 As shown;

[0101] (623) The previous frame F of the current frame -1 The running time t of the system is compared with the single-frame running time t′ required by the real-time performance of the system, and the accuracy s of the straight line detection in the straight line detection is dynamically adjusted; if t>t′, the accuracy s is reduced; if t≤t′, s remains unchanged;

[0102] (624) From the rotation fulcrum K i At the beginning, along the angle θ∈[RP′-min(αV,-σ),RP′+max(αV,σ)], every s pixels are judged according to the gray value of the pixel to see whether it is a continuous straight line, and the length of the straight line at the angle θ is recorded S θ ;

[0103] (625) Repeat step (624) until all angles θ in the angle range [RP′-min(αV,-σ), RP′+max(αV,σ)] are detected, and the longest straight line length S is taken. θ′ and its corresponding angle θ′, with the rotation fulcrum K i Initially, the angle is θ′ and the length is S θ′ As the detection result of linear rigid body i.

[0104] Reference Figure 5 As shown, the method for calculating the reference length in step (7) is specifically as follows:

[0105] According to step (6), the length S of the linear rigid body i in the camera plane in the reference frame is obtained i , use the inertial sensor to obtain the angle k between the linear rigid body i and the camera plane in the reference frame i , obtain the reference length S′ i :

[0106]

[0107] Reference Figure 6 As shown, the calculation method in step (8) is specifically as follows:

[0108] The reference frame T i The deflection angle perpendicular to the camera plane is considered as RS′ i =0; when the length of the linear rigid body i is S i When the deflection angle RS in the direction perpendicular to the camera plane is obtained i :

[0109]

[0110] The present invention has many specific application paths. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principles of the present invention. These improvements should also be considered as the scope of protection of the present invention.

Claims

1. A linear rigid body three-dimensional space motion tracking method based on edge computing, and a linear rigid body three-dimensional space motion tracking system based on edge computing, the system comprising: Remote detection of all-in-one machines, server platforms and clients; The remote detection all-in-one machine is used to obtain real-time linear rigid body motion video stream data, track the linear rigid body's motion in three-dimensional space in the video stream data, obtain the linear rigid body's rotation angle in three-dimensional space, and send the result data to the server platform; The server platform is used to obtain the result data sent by the remote detection machine in real time and store it; The client is used to obtain result data from the server platform in real time and query historical result data; The method is characterized in that it comprises: a detection phase and a tracking phase; In the detection phase, feature information including the number of linear rigid bodies and rotation pivot points is obtained from the video stream data; In the tracking phase, the linear rigid body is detected in a specific area using its characteristic information, and its rotation angle in three-dimensional space is obtained to complete the motion tracking of the linear rigid body in three-dimensional space. The specific steps of the detection stage are: (1) Receive linear rigid body video stream data collected by a camera, and extract the first frame F1 and the second frame F2 from the video stream data; (2) Using the frame difference method, the motion region of the linear rigid body is extracted from the two adjacent frames F1 and F2 to form a mask M; (3) Extract all feature points from the mask M and divide the feature points to obtain the number of linear rigid bodies N' and a set of feature point sets P corresponding to each linear rigid body i i 0 , i∈[1,N']; using linear fitting, we get a distance feature point set P i 0 The shortest straight line between all feature points (4) For each of the subsequent N frames, use the optical flow method to obtain P i 0 The feature point set P matched in the subsequent n-th frame i n , and use linear fitting to get a distance feature point set P i 0 The shortest straight line between all feature points (5) For each linear rigid body, combine N straight lines Fit the nearest intersection point of N straight lines as the rotation fulcrum K of the linear rigid body i i ; The specific steps of the tracking phase are: (6) For the input linear rigid body video stream, use the rotation fulcrum K i The line detection method with a known endpoint detects the rotation angle RP of the linear rigid body i in the camera plane direction i and length S i ; (7) The first frame of the tracking phase is used as the reference frame T of the linear rigid body i i , and using the reference frame T i Calculate the base length of the linear rigid body i as S' i ; (8) The length S of the current frame of the linear rigid body i i and its reference length S′ i For comparison, obtain the rotation angle RS of the linear rigid body in the direction perpendicular to the camera plane i ; (9) Combined with the rotation angle RP of the linear rigid body in the direction of the camera plane i and the rotation angle RS of the linear rigid body in the direction perpendicular to the camera plane i , complete the motion tracking of the linear rigid body i in three-dimensional space.

2. The linear rigid body three-dimensional space motion tracking method based on edge computing according to claim 1 is characterized in that: The feature point division method in step (3) is specifically as follows: (31) Initially, all feature points are in the undivided point set I; (32) Each time a feature point p is taken from the unpartitioned point set I, and a feature point set P is found i , so that P i The center point c i The closest distance to the feature point p, and the distance is less than / equal to the threshold T, the feature point p is put into the feature point set P i If there is no feature point set P that meets the conditions i , then create a new feature point set P n+1 , n is the number of existing feature point sets, and the feature point p is put into the feature point set P n+1 ; Use the Euclidean distance center value of all feature points in the feature point set as the center point c of the feature point set i ; (33) When the undivided point set I is empty, the number of feature point sets obtained is the number of linear rigid bodies N'. Each feature point set represents a feature point set P corresponding to a linear rigid body i. i .

3. The linear rigid body three-dimensional space motion tracking method based on edge computing according to claim 1 is characterized in that: The linear rigid body detection method in step (6) is specifically as follows: (61) Rotate the fulcrum K i As an endpoint of the linear rigid body i; (62) According to the rotation angle RP′ of the linear rigid body in the camera plane direction in the previous frame and the angle threshold r, i As the starting point, a straight line detection with an accuracy of s is performed within the angle range of [RP′-r, RP′+r] to obtain the linear rigid body i; (63) The mask M of the linear rigid body i is obtained by using the difference between the current frame and the previous frame. i , linear rigid body i and mask M i Do the intersection to get the detection results of the linear rigid body i, including the rotation angle RP of the linear rigid body i in the camera plane i and length S i .

4. The linear rigid body three-dimensional space motion tracking method based on edge computing according to claim 3 is characterized in that: The step (62) specifically includes: (621) According to the two frames F before the current frame F0 -2 and F -1 The rotation angles RP″ and RP′ of the linear rigid body in the camera plane direction are used to obtain the current motion velocity V of the linear rigid body i: V = RP′-RP″; (622) According to the motion speed V, the detection angle range is optimized to [RP′-min(αV,-σ),RP′+max(αV,σ)], where α is a hyperparameter used to cope with the change of the linear rigid body's rotation speed; σ is the detection angle accuracy unit used to timely detect the change of the linear rigid body's motion direction; (623) The previous frame F of the current frame -1 The running time t of the system is compared with the single-frame running time t′ required by the real-time performance of the system, and the accuracy s of the straight line detection in the straight line detection is dynamically adjusted; if t>t′, the accuracy s is reduced; if t≤t′, s remains unchanged; (624) From the rotation fulcrum K i At the beginning, along the angle θ∈[RP′-min(αV,-σ),RP′+max(αV,σ)], every s pixels are judged according to the gray value of the pixel to see whether it is a continuous straight line, and the length of the straight line at the angle θ is recorded S θ ; (625) Repeat step (624) until all angles θ in the angle range [RP′-min(αV,-σ),RP′+max(αV,σ)] are detected, and the longest straight line length S is taken. θ′ and its corresponding angle θ′, with the rotation fulcrum K i Initially, the angle is θ′ and the length is S θ′ As the detection result of linear rigid body i.

5. The linear rigid body three-dimensional space motion tracking method based on edge computing according to claim 1 is characterized in that: The method for calculating the reference length in step (7) is specifically as follows: According to step (6), the length S of the linear rigid body i in the camera plane in the reference frame is obtained i , use the inertial sensor to obtain the angle k between the linear rigid body i and the camera plane in the reference frame i , obtain the reference length S′ i :

6. The linear rigid body three-dimensional space motion tracking method based on edge computing according to claim 1, characterized in that: The calculation method in step (8) is specifically as follows: The reference frame T i The deflection angle perpendicular to the camera plane is considered as RS′ i =0; The length of the linear rigid body i is S i When the deflection angle RS in the direction perpendicular to the camera plane is obtained i :

Images