A computer vision-based dual-camera three-stage frog leap testing method and system

Abstract

The application provides a kind of double camera three-stage frog jump test method and system based on computer vision, belong to visual identification technical field. Including: build three-stage frog jump test site, by take-off line is divided into area for preparation area and long jump area, the preparation area and long jump area same side are equipped with one camera respectively;Image captured by camera is regionally calibrated and distortion corrected;Tester enters test area, starts test;Preparation area camera real-time acquisition test image, real-time judge whether there is take-off violation;Tester takes off, start long jump area camera, long jump area camera real-time acquisition test image, judge whether there is jump incoherence violation;Calculate the score of non-violation test, obtain the landing point after removing distortion and take-off line, the vertical distance between the corrected landing point and take-off line is the actual long jump distance. The application unifies the detection standard, reduces the consideration participation guarantee test accuracy and fairness.

Description

Technical Field

[0001] The present invention relates to a dual-camera three-stage frog leaping test method and system based on computer vision, belonging to the technical field of visual recognition. Background Art

[0002] Existing three-level frog jump test methods mainly include the manual three-level frog jump test method and the traditional contact sensor three-level frog jump test method. The manual three-level test method uses a one-to-one or one-to-many ratio of referees to testers. This method has the following problems: high labor costs: multiple referees or a single referee are required to supervise for a long time; inconsistent testing standards: different referees have different standards for judging violations; and difficulty tracing test results: testers cannot review the test process, which makes it impossible to conduct targeted training and trace the source of controversial test results.

[0003] Traditional contact sensor three-stage frog jump test method: The traditional three-stage frog jump test method for contact sensors primarily uses an inductive long jump mat. While this method improves distance measurement accuracy and reduces labor costs by eliminating manual measurement, it still requires human supervision, resulting in high labor costs, inaccurate violation detection, and difficulty in tracing test results.

[0004] In summary, a dual-camera three-level frog leap test method based on computer vision is needed to solve the problems existing in the existing three-level frog leap test method. Summary of the Invention

[0005] The purpose of the present invention is to provide a dual-camera three-level frog leap test method and system based on computer vision, which unifies the detection standards, reduces the number of participants and ensures the accuracy and fairness of the test.

[0006] To achieve the above-mentioned purpose, the present invention is implemented through the following technical solutions:

[0007] Build a three-level frog jump test site, divide the area into a preparation area and a long jump area by the take-off line, and install a camera on the same side of the preparation area and the long jump area respectively;

[0008] Perform regional calibration and distortion correction on images captured by the camera;

[0009] The tester enters the test area, and key data of the tester is collected to determine whether he is ready. If he is ready, the test begins;

[0010] The camera in the preparation area collects test images in real time, detects the test posture of the current tester based on the human key point detection model, obtains the key frames of knee bending, arm swinging, and first take-off, and determines in real time whether there is a take-off violation;

[0011] After the tester takes off, the long jump area camera is turned on. The long jump area camera collects test images in real time and detects the current tester's test posture based on the human key point detection model. The images of the first highest frame in the air, the first take-off landing frame, the second take-off frame, the second highest frame in the air, the second take-off landing frame, the third take-off frame, the third highest frame in the air, and the third take-off landing frame are obtained. The time interval between each jump is calculated to determine whether there is a jump discontinuity violation.

[0012] To calculate the scores of the non-violation tests, the camera in the long jump area detects the tester's third landing and obtains the tester's landing point. Distortion correction is performed on the landing point and take-off line to obtain the landing point and take-off line after dedistortion. The vertical distance from the corrected landing point to the take-off line is the actual long jump distance.

[0013] Preferably, the specific method of area calibration is as follows: mark four points in sequence on the image taken by the camera, where two adjacent points are located on the starting line, and the area connecting the four points is the calibration area, and the two sides of the starting line are the preparation area and the starting area respectively.

[0014] Preferably, the area calibration for the image captured by the camera includes preparation area calibration, take-off line calibration, and long jump area calibration.

[0015] Preferably, the long jump area camera performs distortion correction in the following specific manner:

[0016] Place the checkerboard at multiple locations within the field of view of the camera in the long jump area and take multiple images;

[0017] Call the OpenCV distortion correction function to calculate the intrinsic and extrinsic parameters of the camera, including the camera matrix and distortion coefficients, to eliminate barrel and pincushion distortion in the image;

[0018] According to the four points of the long jump area obtained during calibration and the actual length and width of the long jump area, the perspective transformation matrix of the image is calculated and the image is converted into an orthographic projection form.

[0019] Preferably, the criteria for judging whether the tester is ready are as follows:

[0020] The ankle key point is located in the preparation area. The fluctuation of the ankle key point for 10 consecutive frames is less than the set threshold. When the image resolution is M*N, the threshold is pixel values, the angle formed by the hip key point, the ankle key point and the knee key point is greater than the threshold, the threshold is 160 degrees, and the angle formed by the shoulder key point, the knee key point and the hip key point is greater than the threshold, the threshold is 160 degrees.

[0021] Preferably, the take-off violation includes a line violation, a single-leg take-off violation, a step violation, and a run-up violation.

[0022] The line violation judgment method is as follows:

[0023] The first take-off key frame is obtained, and the current frame is detected through a target detection model to mark the shoes in the image, and it is judged whether the shoe detection box diagram crosses the take-off line. If the take-off line is crossed, it is a line violation;

[0024] The single-leg take-off violation judgment method is as follows:

[0025] The first take-off key frame is obtained, and the distance between the left and right ankle key points is calculated. If the distance is greater than a set threshold value, the threshold value is pixel value when the image resolution is M*N, then the single-leg take-off violation occurs;

[0026] The step violation judgment method is as follows:

[0027] The data of the ankle key points of 5 consecutive frames is obtained. If the ankle key points reach a take-off threshold value in the vertical direction, the threshold value is pixels when the image resolution is M*N, and do not reach a take-off threshold value in the horizontal direction, the threshold value is pixels when the image resolution is M*N, then the step violation occurs;

[0028] The run-up violation judgment method is as follows:

[0029] The data of the ankle key points of 5 consecutive frames is obtained, and the trend of the ankle key points is judged. If the ankle key points gradually move away from the take-off line, then the run-up violation occurs.

[0030] Preferably, the jump incoherence violation judgment method is as follows: the current take-off and landing frame and the next take-off frame are obtained, and the time interval between the last landing and the next take-off is calculated. If the time interval is greater than a set threshold value, the threshold value is 2S, then the jump incoherence violation occurs.

[0031] Preferably, the take-off frame judgment method is as follows:

[0032] The ankle key points of 5 consecutive frames are obtained. If the changes of the ankle key points in the vertical direction and the horizontal direction are both greater than a set threshold value, the threshold value in the vertical direction is pixels when the image resolution is M*N, and the threshold value in the horizontal direction is pixels when the image resolution is M*N, then the current frame is a take-off frame;

[0033] The take-off and landing frame judgment method is as follows:

[0034] Obtain ankle key points for 5 consecutive frames. If the vertical and lateral changes of the ankle key points are both less than the set threshold, and the image resolution is M*N, the threshold is 5 pixels, then the current frame is the take-off and landing frame.

[0035] The key frame determination method is as follows:

[0036] Before taking off, the current knee bending angle is calculated and the minimum knee bending angle before take-off is found, and the current image is the knee bending key frame; before taking off, the current arm swing amplitude is calculated and the maximum arm swing amplitude before take-off is found, and the current frame is the arm swing key frame; when it is determined that the test person takes off, the current test person's take-off angle is calculated, and the current frame is the take-off key frame; after the test person takes off, before the test person lands, the height of the test person's ankle key point is calculated and the highest height is found, and the current frame is the take-off key frame.

[0037] Preferably, the specific method of calculating the score is as follows:

[0038] Perform grayscale conversion on the third take-off and landing frame image through OpenCV, call the OpenCV edge detection algorithm to perform edge detection, and perform inflection point detection on the edge detection result to obtain the landing point of the body;

[0039] The target detection model is called to obtain the location of the tester's shoes, and then image processing is performed on the shoe to obtain the location of the heel of the shoe;

[0040] The obtained landing point and take-off line are simultaneously subjected to distortion correction to obtain the landing point and take-off line after distortion removal. It is determined which landing point is closest to the take-off line. The vertical distance from the corrected landing point to the take-off line is the actual long jump distance.

[0041] A dual-camera three-stage frog leap test system based on computer vision, comprising:

[0042] Image acquisition module: Based on two cameras, it realizes the frame capture function of the entire test process, and obtains frames for real-time intelligent analysis and test process video storage;

[0043] Voice broadcast module: voice broadcast to testers during the test phase;

[0044] Calibration module: used for calibration of the test area under the perspective of two cameras;

[0045] Correction module: used to correct camera distortion, including eliminating barrel distortion, pincushion distortion, and performing perspective transformation on the image to obtain an image under orthographic projection;

[0046] Intelligent test module: based on the human key point detection model in computer vision, the state of the tester is detected, the automatic switching of the double cameras is realized, and the posture analysis, action detection and rule violation judgment of the tester are carried out; based on the shoe detection model and image processing technology, the landing point closest to the take-off line of the tester is obtained; based on distortion correction, the landing point is corrected to obtain the actual long jump distance.

[0047] Communication module: used for communication between various modules, and offline saving of test video and picture.

[0048] The advantages of the present application are that the present application is based on the human key point detection technology, image processing technology and target detection technology in the field of computer vision, realizes the automatic and intelligent camera distortion correction, human action detection, rule violation detection and automatic ranging, unifies the detection standard, reduces the consideration of participation and guarantees the test accuracy and fairness. BRIEF DESCRIPTION OF DRAWINGS

[0049] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, used to explain the present application, and do not constitute a limitation of the present application.

[0050] Figure 1 A flowchart of the three-stage frog jump test method provided by the embodiment of the present application is shown.

[0051] Figure 2 A distortion correction flowchart of the three-stage frog jump test method provided by the embodiment of the present application is shown.

[0052] Figure 3 A top view of the placement position of the camera and the test area in the three-stage frog jump test method provided by the embodiment of the present application is shown. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0054] A double-camera three-stage frog jump test method based on computer vision is realized by the following technical solutions:

[0055] S1: build a three-stage frog jump test site, divide the area into a preparation area and a long jump area through a take-off line, and install one camera on the same side of the preparation area and the long jump area;

[0056] Specifically, a three-level frog jump test site is selected and the entire test site is divided into three parts: preparation area, take-off line and test area. The preparation area is 1m*1m in size, the take-off line is 2m in length, and the long jump area is 9m*1m in size. The installation method is as follows Figure 3 First, install the preparation area camera, level with the take-off line, 3 meters from the preparation area, 2 meters high, with an unobstructed field of view. Adjust the camera angle to encompass the entire preparation area, the take-off line, and part of the test area, and ensure that the tester's entire body is within the camera's field of view when testing in the preparation area. Then, install the test area camera, level with the center line of the long jump area, that is, 4.5 meters from the long jump area, 5 meters from the preparation area, 2.5 meters high, with an unobstructed field of view. Adjust the camera angle to encompass the entire long jump area, and ensure that the tester's entire body is within the camera's field of view when testing in the test area.

[0057] S2: Perform regional calibration and distortion correction on the image captured by the camera; the specific method of regional calibration is as follows:

[0058] Mark the preparation area, take-off line and long jump area, each area is as follows Figure 3 As shown, first, the preparation area camera is turned on and the four points in the preparation area are calibrated. The order of the four points can be set, and the calibration module automatically sorts them according to their relative positions. Then, the take-off line is calibrated in the preparation area camera. The calibration module automatically determines the tester's take-off direction based on the relative position of the take-off line and the preparation area. Finally, the long jump area camera is turned on and the four points in the long jump area are calibrated. The order of the four points can also be set, and the calibration module automatically sorts them according to their relative positions.

[0059] The specific method for performing distortion correction on the long jump area camera is as follows:

[0060] Place the checkerboard at multiple locations within the field of view of the camera in the long jump area and take multiple images;

[0061] Call the OpenCV distortion correction function to calculate the intrinsic and extrinsic parameters of the camera, including the camera matrix and distortion coefficients, to eliminate barrel and pincushion distortion in the image;

[0062] According to the four points of the long jump area and the actual length and width of the long jump area obtained during calibration, the perspective transformation matrix of the image is calculated and the image is converted into an orthographic projection form.

[0063] S3: The tester enters the test area, and key data of the tester is collected to determine whether he is ready. If he is ready, the test begins;

[0064] Specifically, the voice module informs the test personnel to enter the preparation area, and the preparation area camera is turned on. Then, the test personnel enters the preparation area, the camera collects the test personnel image in real time, and the human body key point detection model is used to obtain the human body key points of the test personnel, taking the YOLOv7 model as an example (a total of 17 key points); finally, it is judged that the test personnel is in the preparation area for calibration and is standing ready, and the voice informs the test personnel to start the test.

[0065] The judgment of whether the test personnel is ready is as follows:

[0066] The ankle key point is located in the preparation area, the fluctuation of the ankle key point in the last 10 frames is less than the set threshold value, the image resolution is 1280*720, and the threshold value is 5 pixel values, the included angle formed by the hip key point, the ankle key point and the knee key point is greater than 160 degrees, and the included angle formed by the shoulder key point, the knee key point and the hip key point is greater than 160 degrees.

[0067] S4: The preparation area camera collects test images in real time, detects the test posture of the current test personnel based on the human body key point detection model, obtains the key frames of the first take-off, and judges in real time whether there is a take-off violation; if it is detected that the test personnel steps on the line before take-off, the voice informs the test personnel to step on the line and prepare again, and when the test personnel retreats back to the take-off line, re-enters the test preparation phase.

[0068] The take-off violation includes line stepping violation, single leg take-off violation, step violation and run-up violation;

[0069] The line stepping violation judgment method is as follows:

[0070] The first take-off key frame is obtained, and the current frame is detected by a target detection model. The shoes in the image are marked to judge whether the shoe detection box exceeds the take-off line. If it exceeds the take-off line, it is a line stepping violation;

[0071] The single leg take-off violation judgment method is as follows:

[0072] The first take-off key frame is obtained, and the distance between the left and right ankle key points is calculated. If the distance is greater than the set threshold value, the image resolution is 1280*720, and the threshold value is 45 pixel values, then the single leg take-off violation occurs;

[0073] The step violation judgment method is as follows:

[0074] The data of the ankle key points of the last 5 frames is obtained. If the ankle key points reach the take-off threshold value in the vertical direction, the image resolution is 1280*720, and the threshold value is 15 pixel values, and the ankle key points do not reach the take-off threshold value in the horizontal direction, the image resolution is 1280*720, and the threshold value is 15 pixel values, then the step violation occurs;

[0075] The method for judging the violation of the run-up is as follows:

[0076] Obtain 5 consecutive frames of ankle key point data to determine the changing trend of the ankle key points. If the ankle key points gradually move away from the take-off line, the run-up is illegal.

[0077] S5: After the tester takes off, the long jump area camera is turned on. The long jump area camera collects test images in real time, and detects the current tester's test posture based on the human key point detection model. The images of the first highest frame in the air, the first take-off landing frame, the second take-off frame, the second take-off highest frame in the air, the second take-off landing frame, the third take-off frame, the third highest frame in the air, and the third take-off landing frame are obtained. The time interval between each jump is calculated to determine whether there is a jump discontinuity violation.

[0078] The jump incoherence violation is determined as follows: the current take-off and landing frame and the next take-off frame are obtained, and the time interval between the last landing and the next take-off is calculated. If the time interval is greater than 2S, the jump incoherence violation occurs.

[0079] Specifically, the jump frame determination method is as follows:

[0080] Get ankle key points for five consecutive frames. If the vertical and lateral changes of the ankle key points are greater than the set threshold, which is 15 pixels when the image resolution is 1280*720, the current frame is the jump frame.

[0081] The take-off and landing frames are determined as follows:

[0082] Get 5 consecutive frames of ankle key points. If the vertical and lateral changes of the ankle key points are less than the set threshold, and the image resolution is 1280*720, the value is 5 pixels, then the current frame is the take-off and landing frame.

[0083] S6: Calculate the score of the non-violation test. After the long jump area camera detects the tester's third landing, obtain the tester's landing point, perform distortion correction on the landing point and take-off line, and obtain the landing point and take-off line after dedistortion. The vertical distance from the corrected landing point to the take-off line is the actual long jump distance.

[0084] The specific method of calculating the scores is as follows:

[0085] Perform grayscale conversion on the third take-off and landing frame image through OpenCV, call the OpenCV edge detection algorithm to perform edge detection, and perform inflection point detection on the edge detection result to obtain the landing point of the body;

[0086] The target detection model is called to obtain the location of the tester's shoes, and then image processing is performed on the shoe to obtain the location of the heel of the shoe;

[0087] The obtained landing point and take-off line are simultaneously subjected to distortion correction to obtain the landing point and take-off line after distortion removal. It is determined which landing point is closest to the take-off line. The vertical distance from the corrected landing point to the take-off line is the actual long jump distance.

[0088] Example 2

[0089] A dual-camera three-stage frog leap test system based on computer vision, comprising:

[0090] Image acquisition module: Based on two cameras, it realizes the frame capture function of the entire test process, and obtains frames for real-time intelligent analysis and test process video storage;

[0091] Voice broadcast module: voice broadcast to testers during the test phase;

[0092] Calibration module: used for calibration of the test area under the perspective of two cameras;

[0093] Correction module: used to correct camera distortion, including eliminating barrel distortion, pincushion distortion, and performing perspective transformation on the image to obtain an image under orthographic projection;

[0094] Intelligent testing module: Based on the human key point detection model in computer vision, it realizes tester status detection, automatic switching of dual cameras, and performs posture analysis, motion detection and violation judgment on the tester; based on the shoe detection model and image processing technology, it obtains the tester's landing point closest to the take-off line; based on distortion correction, it corrects the landing point to obtain the actual long jump distance.

[0095] Communication module: used for communication between modules and to save test videos and pictures offline.

[0096] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A dual-camera three-level frog leap test method based on computer vision, characterized in that: include: Build a three-level frog jump test site, divide the area into a preparation area and a long jump area by the take-off line, and install a camera on the same side of the preparation area and the long jump area respectively; Perform regional calibration and distortion correction on images captured by the camera; The tester enters the test area, and key data of the tester is collected to determine whether he is ready. If he is ready, the test begins; The camera in the preparation area collects test images in real time, detects the test posture of the current tester based on the human key point detection model, obtains the key frames of knee bending, arm swinging, and first take-off, and determines in real time whether there is a take-off violation; After the tester takes off, the long jump area camera is turned on. The long jump area camera collects test images in real time and detects the current tester's test posture based on the human key point detection model. The images of the first highest frame in the air, the first take-off landing frame, the second take-off frame, the second highest frame in the air, the second take-off landing frame, the third take-off frame, the third highest frame in the air, and the third take-off landing frame are obtained. The time interval between each jump is calculated to determine whether there is a jump discontinuity violation. To calculate the score for a non-violation test, the camera in the long jump area detects the tester's third landing and obtains the tester's landing point. Distortion correction is performed on the landing point and take-off line to obtain the dedistorted landing point and take-off line. The vertical distance from the corrected landing point to the take-off line is the actual long jump distance. The specific method for performing distortion correction on the long jump area camera is as follows: Place the checkerboard at multiple locations within the field of view of the camera in the long jump area and take multiple images; Call the OpenCV distortion correction function to calculate the intrinsic and extrinsic parameters of the camera, including the camera matrix and distortion coefficient, to eliminate barrel and pincushion distortion in the image; According to the four points of the long jump area obtained during calibration and the actual length and width of the long jump area, the perspective transformation matrix of the image is calculated and the image is converted into an orthographic projection form; The criteria for judging whether a tester is ready are as follows: The ankle key point is located in the preparation area. The fluctuation of the ankle key point for 10 consecutive frames is less than the set threshold. When the image resolution is M*N, the threshold is pixel values, the angle formed by the hip key point, the ankle key point and the knee key point is greater than the threshold, the threshold is 160 degrees, the angle formed by the shoulder key point, the knee key point and the hip key point is greater than the threshold, the threshold is 160 degrees; The aforementioned take-off violations include line-pressing violations, single-foot take-off violations, step-over violations, and run-up violations; The method for judging the violation of the line pressure is as follows: Get the first take-off keyframe and detect the current frame using the target detection model. Mark the shoes in the image and determine whether the shoe detection frame crosses the take-off line. If so, it is a line violation. The method for judging the violation of the single-leg jump is as follows: Get the first jump key frame and calculate the distance between the left and right ankle key points. If the distance is greater than the set threshold, the threshold is pixel value, then the single-foot jump is illegal; The method for judging the violation of the above-mentioned stepping is as follows: Get 5 consecutive frames of ankle key point data. If the ankle key point reaches the take-off threshold in the vertical direction, when the image resolution is M*N, the threshold is pixel values, the horizontal change does not reach the jump threshold, when the image resolution is M*N, the threshold is pixel values, then the step is illegal; The method for judging the violation of the run-up is as follows: Obtain 5 consecutive frames of ankle key point data to determine the changing trend of the ankle key points. If the ankle key points gradually move away from the take-off line, the run-up is illegal.

2. The dual-camera three-stage frog leap test method based on computer vision according to claim 1, characterized in that: The specific method of area calibration is as follows: mark four points in sequence on the image taken by the camera, where two adjacent points are located on the starting line. The area connected by the four points is the calibration area, and the two sides of the starting line are the preparation area and the starting area respectively.

3. The dual-camera three-stage frog leap test method based on computer vision according to claim 1, characterized in that: The regional calibration for the image captured by the camera includes preparation area calibration, take-off line calibration, and long jump area calibration.

4. The dual-camera three-stage frog leap test method based on computer vision according to claim 1, characterized in that: The jump incoherence violation is determined as follows: the current take-off and landing frame and the next take-off frame are obtained, and the time interval between the last landing and the next take-off is calculated. If the time interval is greater than a set threshold, the threshold is 2S, then the jump incoherence violation occurs.

5. The dual-camera three-stage frog leap test method based on computer vision according to claim 1, characterized in that: The jump frame is determined as follows: Get 5 consecutive frames of ankle key points. If the changes of the ankle key points in the vertical and lateral directions are greater than the set threshold, when the image resolution is M*N, the vertical threshold is pixel values, the horizontal threshold is pixel values, the current frame is the jump frame; The take-off and landing frames are determined as follows: Obtain ankle key points for five consecutive frames. If the vertical and lateral changes of the ankle key points are both less than a set threshold, and the image resolution is M*N, the threshold is 5 pixels, then the current frame is a take-off and landing frame. The key frame determination method is as follows: Before taking off, the current knee bending angle is calculated and the minimum knee bending angle before take-off is found, and the current image is the knee bending key frame; before taking off, the current arm swing amplitude is calculated and the maximum arm swing amplitude before take-off is found, and the current frame is the arm swing key frame; when it is determined that the test person takes off, the current test person's take-off angle is calculated, and the current frame is the take-off key frame; after the test person takes off and before the test person lands, the height of the test person's ankle key point is calculated and the highest height is found, and the current frame is the take-off key frame.

6. The dual-camera three-stage frog leap test method based on computer vision according to claim 1, characterized in that: The specific method of calculating the scores is as follows: Perform grayscale conversion on the third take-off and landing frame image through OpenCV, call the OpenCV edge detection algorithm to perform edge detection, and perform inflection point detection on the edge detection result to obtain the landing point of the body; The target detection model is called to obtain the location of the tester's shoes, and then image processing is performed on the shoe to obtain the location of the heel of the shoe; The obtained landing point and take-off line are simultaneously subjected to distortion correction to obtain the landing point and take-off line after distortion removal. It is determined which landing point is closest to the take-off line. The vertical distance from the corrected landing point to the take-off line is the actual long jump distance.

7. A dual-camera three-stage frog leap test system based on computer vision using any one of claims 1-6, characterized in that: include: Image acquisition module: Based on two cameras, it realizes the frame capture function of the entire test process, and obtains frames for real-time intelligent analysis and test process video storage; Voice broadcast module: voice broadcast to testers during the test phase; Calibration module: used for calibration of the test area under the perspective of two cameras; Correction module: used to correct camera distortion, including eliminating barrel distortion, pincushion distortion, and performing perspective transformation on the image to obtain an image under orthographic projection; Intelligent testing module: Based on the human key point detection model in computer vision, it realizes the tester's status detection, automatic switching of dual cameras, and performs posture analysis, motion detection and violation judgment on the tester. It also uses the shoe detection model and image processing technology to obtain the tester's landing point closest to the take-off line. Based on distortion correction, the landing point is corrected to obtain the actual long jump distance; Communication module: used for communication between modules and to save test videos and pictures offline.

Images