Pickling action recognition calculation method
Through deep learning single-stage object detection algorithm and human skeleton key point detection, the problem that the existing technology cannot effectively judge soldiers' stabbing movements, realize the accurate identification of stabbing movements and provide training suggestions, and improve the stabbing training and assessment. Scientific and effective.
Patent Information
- Application Number
- CN202411095830.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing sprint judgment system cannot effectively judge the basic sprint movements of soldiers and cannot achieve scientific, accurate and efficient training and assessment.
Deep learning single-stage object detection algorithm combined with top-down detection of key points of human bones, the camera takes the stinging movements of trainees, and detects key points such as nose, eyes, ears, shoulders, elbows, wrists, hips, knees, ankles. , perform spatial geometry calculations, judge the movement characteristics such as kicking distance, arm straightening, calf vertical, and body oblique straightening, calculate the number of effective spikes, and feedback the results to the client.
It realizes accurate identification of soldiers' stabbing movements and collection of wrong movements, provides training suggestions, and improves the scientificity and effectiveness of stabbing training and assessment.
Smart Images

Figure CN119049122A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a stabbing action recognition calculation method. Background Art
[0002] The thrust evaluation system is a product of the current development trend of the army. It aims to realize the scientificization of daily training and military assessment, dataization of processes, and visualization of effects for the troops, so as to comprehensively improve the basic physical fitness level of soldiers, improve military skills and combat effectiveness. It mainly uses video perception, recognition technology, and intelligent communication technology to establish a scientific, accurate, efficient, and intelligent training and assessment system. However, the existing technology cannot effectively judge the basic thrust actions of soldiers. For this reason, we propose a stabbing action recognition calculation method. Summary of the invention
[0003] The main purpose of the present invention is to provide a stabbing action recognition calculation method, which can effectively solve the problems in the background technology.
[0004] To achieve the above object, the technical solution adopted by the present invention is:
[0005] The stabbing action recognition calculation method includes the following steps:
[0006] S1. The camera is used to shoot the trainees stabbing the human-shaped target. The individual stabbing recognition service system accesses the real-time video stream through the router in the RTSP protocol mode and detects the key frames in the video stream.
[0007] S2. During detection, a single-stage deep learning target detection algorithm is used to detect that the target holding a gun and in a ready-to-use gun posture is determined to be a trainee;
[0008] S3, track the trainees frame by frame, and use top-down human skeleton key point detection for key actions, detecting 17 key points including nose, left and right eyes, left and right ears, left and right shoulders, left and right elbows, left and right wrists, left and right hips, left and right knees, and left and right ankles;
[0009] S4. For the detected key points, a spatial geometric calculation is performed with a human-shaped target as a reference system. When the system defaults to right-handedness, the left kicking distance, whether the left arm is straight, whether the left calf is perpendicular to the ground, and whether the body is in an oblique straight line are obtained. These detection results are fed back in the form of an interface, and the corresponding effective thrust number is obtained at the same time;
[0010] S5. The individual soldier breakthrough pushes the recognition result to the client through the router, and the stabbing pressure value of the human target pushes the data to the client through the router.
[0011] Furthermore, in S4, the kicking distance is the distance between the position of the left foot of the trainee when preparing for the thrust and the position of the left foot after the thrust is kicked out and lands. The algorithm for the kicking distance determines that the foot must be kicked out about 5° and the kicking angle is about 15-25 cm.
[0012] Furthermore, in S4, the arm straightening means that the trainee's left arm is straightened when the trainee stabs the target with a thrust action, and the arm straightening algorithm determines that the gun face is rotated about 5°.
[0013] Furthermore, in S4, the calf being perpendicular to the ground means that after the trainee's left foot kicks out and lands on the ground in a lunging action, the calf is perpendicular to the ground. The algorithm determines that the calf is perpendicular to the ground within 15° backward or forward, otherwise an error occurs.
[0014] Furthermore, in S4, the body is straight and diagonal, which means that the trainee stabs the target with a thrusting action, and after the left foot lands on the ground, the body is straight and diagonal, wherein the body straight line algorithm determines that the body cannot lean back.
[0015] Furthermore, in S4, an effective thrust means that a trainee's thrust action hits a target, which is an effective thrust. An effective thrust action recognition means that a trainee completes an effective thrust action, and the algorithm recognizes the thrust action, which is an effective thrust action recognition. The thrust action recognition rate = the number of effective thrust action recognitions / the number of effective thrust actions completed by the trainee*100%.
[0016] Compared with the prior art, the present invention has the following beneficial effects:
[0017] In the present invention, the stabbing action recognition calculation method can recognize the soldier's action, collect erroneous actions, and give training suggestions, thereby realizing the action soldiers of the stabbing training and assessment, and judging whether the soldier's action reaches the standard. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is a functional flow chart of the assassination action recognition calculation method of the present invention;
[0019] Figure 2 This is a service structure diagram of the assassination action recognition calculation method of the present invention. DETAILED DESCRIPTION
[0020] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.
[0021] Embodiment 1:
[0022] like Figure 1-2As shown in the figure, the assassination action recognition calculation method includes the following steps: S1, using a camera to shoot the trainee stabbing a human-shaped target, the single-soldier stabbing recognition service system accesses the real-time video stream through the router in the RTSP protocol mode, and detects the key frames in the video stream; S2, during the detection, a deep learning single-stage target detection algorithm is used to detect the target holding a gun and in a ready-to-use gun posture as a trainee; S3, the trainee is tracked frame by frame, and the key actions are detected by using the top-down human skeleton key Point detection, detects 17 key points including nose, left and right eyes, left and right ears, left and right shoulders, left and right elbows, left and right wrists, left and right hips, left and right knees, and left and right ankles; S4, for the detected key points, uses the human-shaped target as the reference system for spatial geometric calculations. Under the default right-handed system, the left kicking distance, whether the left arm is straight, whether the left calf is perpendicular to the ground, and whether the body is in a straight line are obtained. These detection results are fed back in the form of an interface, and the corresponding effective number of thrusts is obtained at the same time. The kicking distance is the distance between the left foot position of the trainee when preparing for the thrust and the thrust kick. The distance of the left foot after landing, the algorithm for kicking distance is that the foot must be kicked out about 5°, and the kicking distance is about 15-25 cm. The arm is straight, which means that the left arm of the trainee is straight when the target is stabbed with a thrust. The algorithm for straightening the arm is that the gun is turned about 5°. The calf is perpendicular to the ground, which means that after the trainee's left foot kicks out and lands, the calf is perpendicular to the ground. The algorithm for perpendicularity between the calf and the ground is within 15° backward and forward. If it exceeds this, it will be a mistake. The body is straight, which means that after the trainee stabs the target with a thrust, the body is straight after the left foot lands. Among them, the algorithm judges that the body cannot lean back, effective stab means that the trainee stabs the target with a stab action, which is an effective stab, and effective stab action recognition means that the trainee completes an effective stab action, and the algorithm recognizes the stab action, which is an effective stab action recognition, and the stab action recognition rate = the number of effective stab action recognitions / the number of effective stab actions completed by the trainee*100%; S5, the individual breakthrough pushes the recognition result to the client through the router, and the stab pressure value of the human-shaped target pushes the data to the client through the router;
[0023] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. The assassination action recognition calculation method includes the following steps: S1. The camera is used to shoot the trainees stabbing the human-shaped target. The individual stabbing recognition service system accesses the real-time video stream through the router in the RTSP protocol mode and detects the key frames in the video stream. S2. During detection, a single-stage deep learning target detection algorithm is used to detect that the target holding a gun and in a ready-to-use gun posture is determined to be a trainee; S3, track the trainees frame by frame, and use top-down human skeleton key point detection for key actions, detecting 17 key points including nose, left and right eyes, left and right ears, left and right shoulders, left and right elbows, left and right wrists, left and right hips, left and right knees, and left and right ankles; S4. For the detected key points, a spatial geometric calculation is performed with a human-shaped target as a reference system. When the system defaults to right-handedness, the left kicking distance, whether the left arm is straight, whether the left calf is perpendicular to the ground, and whether the body is in an oblique straight line are obtained. These detection results are fed back in the form of an interface, and the corresponding effective thrust number is obtained at the same time; S5. The individual soldier breakthrough pushes the recognition result to the client through the router, and the stabbing pressure value of the human target pushes the data to the client through the router.
2. The stabbing action recognition calculation method according to claim 1, characterized in that: In S4, the kicking distance is the distance between the position of the left foot of the trainee when preparing for the thrust and the position of the left foot after the thrust is kicked out and lands. The algorithm for the kicking distance determines that the foot must be kicked out about 5°, and the kicking angle is about 15-25 cm.
3. The stabbing action recognition calculation method according to claim 1, characterized in that: In S4, the arm straightening means that the trainee's left arm is straightened when the trainee stabs the target with a thrust action, and the arm straightening algorithm determines that the gun face is rotated about 5°.
4. The stabbing action recognition calculation method according to claim 1, characterized in that: In S4, the calf being perpendicular to the ground means that after the trainee's left foot kicks out and lands on the ground during a lunging action, the calf is perpendicular to the ground. The algorithm determines that the calf is perpendicular to the ground within 15° backward or forward, otherwise an error occurs.
5. The assassination action recognition calculation method according to claim 1, characterized in that: In S4, the body is straight and diagonal, which means that the trainee stabs the target with a thrust action, and after the left foot lands on the ground, the body is straight and diagonal, wherein the body straight and diagonal algorithm determines that the body cannot lean back.
6. The assassination action recognition calculation method according to claim 1 is characterized in that: In S4, an effective thrust means that the trainee's thrust action hits the target, which is an effective thrust. An effective thrust action recognition means that the trainee completes an effective thrust action, and the algorithm recognizes the thrust action, which is an effective thrust action recognition. The thrust action recognition rate = the number of effective thrust action recognitions / the number of effective thrust actions completed by the trainee*100%.
