Hemispherical prone push-up arm measuring method and implementation device

Abstract

The application discloses a kind of hemispherical prone push-up arm measurement method and implementation device, it is related to physical training and physical testing technical field, including the following steps: the calibration of mark line is carried out;Receive a frame image;Carry out human key point identification;Judge working state;Set working state as counting state;Detect whether the tester leaves;Detect whether body touches ground;Update touch ground counter;Detect whether hand is in the left of first mark line;Update the counter of hand in the left of first mark line;Judge whether hand has passed the right of second mark line;Judge whether body has touched ground;Reset action counting related parameters;Detect whether hand is in the right of second mark line;Update the counter of hand in the right of second mark line;Test end.Through the number of action based on the 2D coordinate calculation of human key point, whether it is detected to be guilty, while entire test process is executed in real time, provides real-time feedback to tester, is corrected to the improper action, to reach the best test result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of physical training and physical testing, in particular to a method for measuring semispherical push-up and arm row and an implementation device thereof. BACKGROUND

[0002] With the development of society, people's demand for their own physical fitness level is getting higher and higher. Not only do they want to have a healthy body, but also a strong physique and good balance ability. Therefore, it is necessary to accurately understand their own physical fitness level through objective measurement methods.

[0003] There are two existing measurement methods: 1) measurement by manual observation, and 2) measurement by infrared sensor. The measurement and counting by manual observation are greatly affected by the measurement staff, and it is difficult to maintain a unified judgment standard for all measured persons, which may lead to unfairness and inaccuracy. The measurement method by infrared sensor is complex to install, and the signal generating device and the signal receiving device must be aligned. If there is any deviation, the measurement cannot be performed. In addition, this method can only detect the set points, and has no knowledge about the physical state of the measured person, and cannot judge the basic movement requirements (such as whether the body shakes, whether the hands and feet touch the ground, etc.). SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a method for measuring semispherical push-up and arm row and an implementation device thereof to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solution: a method for measuring semispherical push-up and arm row and an implementation device thereof, comprising a stand and a test area, wherein the bottom end of the stand is fixedly installed with a stand base, and the top end of the stand is fixedly installed with a smart screen on one side surface.

[0006] A semisphere is arranged on the test area, and a first mark line and a second mark line are arranged on one side of the semisphere.

[0007] Further optimization of the technical solution is that the installation position of the smart screen is opposite to the test area, the long side of the test area is parallel to the long side of the smart screen, the smart screen is internally provided with a camera, and the camera is internally provided with a human body key point recognition model.

[0008] Further optimization of the technical solution is that the first mark line and the second mark line are arranged in parallel to each other, and both mark lines are perpendicular to the long side of the test area, and the distance between the first mark line and the second mark line is 66 cm.

[0009] A method for measuring semispherical push-up and arm row, based on the implementation device described above, comprises the following steps:

[0010] S1, calibrate the marking line;

[0011] S2, start the test: configure and start the camera, configure and start the human key point detection model, and then initialize the state variable;

[0012] S3, receive a frame of image: read a frame of 1920x1080 MJPEG image from the camera, decompress it into YUV420 format image, and finally send it to the human key point recognition model;

[0013] S4, perform human key point recognition;

[0014] S5, judge the working state: if the current working state is detection, execute step S6, if the current working state is counting, execute step S8;

[0015] S6, detect whether the tester is ready for the test;

[0016] S7, set the working state to the counting state, and execute step S3;

[0017] S8, detect whether the tester has left: if there is no human body frame or the human body frame does not meet the requirements, it means that the tester has left, then execute step S20, otherwise execute step S9;

[0018] S9, detect whether the body has touched the ground: if the body has touched the ground, execute step S10, otherwise execute step S11;

[0019] S10, update the touch ground counter;

[0020] S11, detect whether the hand is on the left side of the first marking line;

[0021] S12, update the counter of the hand on the left side of the first marking line;

[0022] S13, judge whether the hand has passed the right side of the second marking line;

[0023] S14, judge whether the body has touched the ground;

[0024] S15, perform error counting;

[0025] S16, perform correct counting;

[0026] S17, reset the action counting related parameters;

[0027] S18, detect whether the hand is on the right side of the second marking line;

[0028] S19, update the counter of the hand on the right side of the second marking line;

[0029] S20, end the test.

[0030] Further optimize the technical solution, in the step S1, the calibration of the mark line includes:

[0031] After the implementation device of the test is installed, the intelligent screen displays the image shot by the camera in real time during calibration, and there are four movable points on the screen. The four points are moved to the four endpoints of the two mark lines respectively, and then the confirm button is clicked.

[0032] After clicking the confirm button, the 2D coordinates of the four endpoints are obtained. The two points with smaller x coordinates (x1, y1) and (x2, y2) determine the first mark line, and the two points with larger x coordinates (x3, y3) and (x4, y4) determine the second mark line (6). Then the straight line equation corresponding to the first mark line is equation one, and the straight line equation corresponding to the second mark line is equation two.

[0033] Equation one and equation two are as follows:

[0034] (y2-y1)x+(x1-x2)y+(x2y1-x1y2)=0 equation one.

[0035] (y4-y3)x+(x3-x4)y+(x4y3-x3y4)=0 equation two.

[0036] Further optimize the technical solution, in the step S4, the pre-processing of the human key point recognition model first down-samples the 1920x1080 YUV420 image to a 960x540 image, and then feeds it to the human key point recognition model. After feature extraction, key point detection and post-processing of the human key point recognition model, the 19 2D key point coordinate positions and their confidence (x i ,y i ,c i ) of the human body are output, (x i ,y i ) is the coordinate of the i-th key point, (c i ) is the confidence of the i-th key point, and the human body frame BBoc(x, y, w, h, c) is output, (x, y) is the left upper corner coordinate of the human body frame, (w, h) is the width and height of the human body frame, and (c) is the confidence of the human body frame.

[0037] Further optimize the technical solution, in the step S6, the test process includes:

[0038] First step: find the tester. Since the test action is prone, the rule for finding the tester is:

[0039] BBox.w>BBox.h and BBox.w>min_width and BBox.c>confidence_thres

[0040] that is, the width of the human body frame is greater than the height of the human body frame, the width of the human body frame is greater than the parameter min_width, and the confidence of the human body frame is greater than the parameter confidence_thres;

[0041] If multiple human body frames satisfying the above rules are detected, the human body frame with the maximum value of (BBox.y+BBox.h) is taken, that is, the human body frame closest to the smart screen;

[0042] The second step is to detect whether the tester is ready, and the judgment conditions include that the hand is on the left side of the first mark line and the body does not touch the ground.

[0043] Further optimize the technical solution, in the step S8, the human body frame or the human body frame does not satisfy the requirements as follows:

[0044] BBox.w>BBox.h and BBox.w>min_width

[0045] Further optimize the technical solution, in the step S10, when updating the touch ground counter, if the body touches the ground, the value of touchGroud is added by 1, as shown in the following formula, and then the step S11 is executed.

[0046] touchGround += 1

[0047] Further optimize the technical solution, in the step S20, when detecting that the tester leaves, it indicates that the test is over, the correct number correctNum and the error number errNum are displayed on the interface, after 10 seconds of display, the following data is reset, and then the step S3 is executed.

[0048] Working state = detection

[0049] fiveLeft = 0; sixRight = 0;

[0050] touchGround = 0;

[0051] correctNum = 0; errNum = 0;

[0052] Compared with the prior art, the present application provides a kind of hemispherical prone push-up arm measurement method and realizing device, with the following beneficial effects:

[0053] The hemisphere prone push-up arm row measurement method and the implementation device thereof calculate the number of actions based on the 2D coordinates of the human body key points, detect whether the rule is violated, and simultaneously perform the whole test process in real time, and the processing speed is higher than 25 frames per second, which provides real-time feedback for the tester, so that the user can know the test state in real time, and the improper action is corrected in time to achieve the best test result. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 A structural schematic view of an implementation device for measuring a hemisphere prone push-up arm row is provided for the present application.

[0055] Figure 2 A flowchart of a hemisphere prone push-up arm row measurement method is provided for the present application.

[0056] In the figure: 1, stand; 2, stand base; 3, smart screen; 4, hemisphere; 5, first mark line; 6, second mark line; 7, test area. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with 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 fall within the scope of protection of the present application.

[0058] Embodiment one:

[0059] Please refer to Figure 1 An implementation device for measuring a hemisphere prone push-up arm row includes a stand 1 and a test area 7, and the bottom end of the stand 1 is fixedly installed with a stand base 2, and one side surface of the top end of the stand 1 is fixedly installed with a smart screen 3.

[0060] Among them, the test area 7 is provided with a hemisphere 4, and one side of the hemisphere 4 is respectively provided with a first mark line 5 and a second mark line 6 located on the test area 7.

[0061] The installation position of the smart screen 3 is opposite to the test area 7, the long side of the test area 7 is parallel to the long side of the smart screen 3, the smart screen 3 is built-in with a 1080P camera, and the camera is built-in with a human body key point recognition model. That is, the smart screen 3 can output the 2D coordinates of the human body key points.

[0062] Among them, the first mark line 5 and the second mark line 6 are arranged in parallel to each other, and both mark lines are perpendicular to the long side of the test area 7, and the distance between the first mark line 5 and the second mark line 6 is 66 centimeters.

[0063] Embodiment two:

[0064] Referring to Figure 2 A method for measuring a semisupine roll and arm swing, based on the implementation device of embodiment one, comprising the following steps:

[0065] S1, calibrate the mark line.

[0066] In this step, after the implementation device is installed, the smart screen 3 displays the image captured by the camera in real time during calibration, and there are four movable points on the screen. Move the four points to the four endpoints of the two mark lines respectively, and then click the confirm button.

[0067] After clicking confirm, the 2D coordinates of the four endpoints are obtained, the two points with smaller x coordinates (x1, y1) and (x2, y2) determine the first mark line 5, and the two points with larger x coordinates (x3, y3) and (x4, y4) determine the second mark line 6. Then the straight line equation corresponding to the first mark line 5 is equation one, and the straight line equation corresponding to the second mark line 6 is equation two.

[0068] Equation one and equation two are as follows:

[0069] (y2-y1)x+(x1-x2)y+(x2y1-x1y2)=0 equation one;

[0070] (y4-y3)x+(x3-x4)y+(x4y3-x3y4)=0 equation two.

[0071] S2, start testing: configure and start the camera, configure and start the human body key point detection model, and then initialize the state variable.

[0072] The initialized state variable is as follows:

[0073] Working state = detection

[0074] fiveLeft = 0

[0075] sixRight = 0

[0076] touchGround = 0

[0077] correctNum = 0

[0078] errNum = 0

[0079] thres1 = 3

[0080] thres2 = 3

[0081] min_width = 400

[0082] confidence_thres = 0.8

[0083] touch_thres = 20

[0084] S3, receiving a frame of image: reading a frame of 1920x1080 MJPEG image from the camera, and decompressing into YUV420 format image, and finally sending to the human key point recognition model.

[0085] S4, human key point recognition.

[0086] In this step, the pre-processing of the human key point recognition model first down-samples the 1920x1080 YUV420 image into a 960x540 image, and then gives it to the human key point recognition model. After feature extraction, key point detection and post-processing of the human key point recognition model, the 19 2D key point coordinate positions and their confidence (x i , y i , c i ) of the human body are outputted, (x i , y i ) is the coordinate of the i-th key point, (c i ) is the confidence of the i-th key point.

[0087] The names and serial numbers of the 19 key points are:

[0088] Nose (0), left eye (1), right eye (2), left ear (3), right ear (4), left shoulder (5), right shoulder (6), left elbow (7), right elbow (8), left wrist (9), right wrist (10), left hand (11), right hand (12), left hip (13), right hip (14), left knee (15), right knee (16), left ankle (17), right ankle (18).

[0089] And output the human body frame BBox(x, y, w, h, c), (x, y) is the left upper corner coordinate of the human body frame, (w, h) is the width and height of the human body frame, (c) is the confidence of the human body frame.

[0090] S5, judging the working state: if the current working state is detection, then executing step S6, if the current working state is counting, then executing step S8.

[0091] S6, detecting whether the tester is ready for the test.

[0092] In this step, the test process includes:

[0093] First step: find the tester, since the test action is prone, so the rule for finding the tester is:

[0094] BBox.w>BBox.h and BBox.w>min_width and BBox.c>confidence_thres

[0095] That is, the width of the human body frame is greater than the height of the human body frame, and the width of the human body frame is greater than the parameter min_width, and the confidence of the human body frame is greater than the parameter confidence_thres;

[0096] If multiple human body frames satisfying the above rules are detected, take the human body frame with the maximum value of (BBox.y+BBox.h), that is, the human body frame closest to the smart screen 3;

[0097] Second step: detect whether the tester is ready, the judgment conditions include: 1) hand is on the left side of the first mark line 5 and 2) body does not touch the ground.

[0098] Determine whether both hands are on the left side of the first mark line 5: substitute the 2D coordinates of leftHand and rightHand into equation (1), if the results are both less than 0, then both hands are on the left side of the first mark line 5.

[0099] Determine whether the body touches the ground: since leftHip and rightHip are located above the hemisphere, these two parts will not touch the ground, and other points compared with them within the set threshold touch_thres range, then no touch ground, otherwise touch ground.

[0100] leftHand.y-leftHip.y<touch_thres

[0101] leftElbow.y-leftHip.y<touch_thres

[0102] leftKnee.y-leftHip.y<touch_thres

[0103] leftAnkle.y-leftHip.y<touch_thres / 2

[0104] rightHand.y-rightHip.y<touch_thres

[0105] rightElbow.y-rightHip.y<touch_thres

[0106] rightKnee.y - rightHip.y < touch_thres

[0107] rightAnkle.y - rightHip.y < touch_thres / 2

[0108] If the above conditions are all met, it means the body is not touching the ground, otherwise the body has touched the ground.

[0109] When 1) both hands are detected to be on the left of the first marker line 5, and 2) the body is detected not to touch the ground, it means the tester is ready, then step S7 is executed, otherwise step S3 is executed.

[0110] S7, set the working state to the counting state, and execute step S3.

[0111] S8, detect whether the tester has left: if there is no human body frame or the human body frame does not meet the requirements, it means the tester has left, then step S20 is executed, otherwise step S9 is executed.

[0112] In this step, the human body frame or the human body frame not meeting the requirements is as follows:

[0113] BBox.w > BBox.h and BBox.w > min_width

[0114] S9, detect whether the body touches the ground: if the body is detected to touch the ground, step S10 is executed, otherwise step S11 is executed.

[0115] S10, update the touch ground counter.

[0116] In this step, when updating the touch ground counter, if the body is detected to touch the ground, the value of touchGround is added by 1, as shown in the following formula, and then step S11 is executed.

[0117] touchGround += 1

[0118] S11, detect whether the hands are on the left of the first marker line 5.

[0119] In this step, the detection method is consistent with the method in 1) of step S6, if both hands are on the left of the first marker line 5, step S12 is executed, otherwise step S18 is executed.

[0120] S12, update the counter of the hands on the left of the first marker line 5.

[0121] When the hands are detected to be on the left of the first marker line 5, the counter of the hands on the left of the first marker line 5 is added by 1, and then step S13 is executed.

[0122] fiveLeft += 1

[0123] S13, determine whether the hands have crossed the right side of the second marker line 6.

[0124] If the counter sixRight of the hands crossing the right side of the second marker line 6 is greater than the threshold value thresl, it is determined that the hands have crossed the right side of the second marker line 6, and counting is performed, and step S14 is executed; otherwise, counting is not performed, and step S3 is executed.

[0125] S14, determine whether the body has touched the ground.

[0126] If the counter touchGround of the body touching the ground is greater than the threshold value thres2, it is determined that the body has touched the ground, and error counting is performed, and step S15 is executed; otherwise, correct counting is performed, and step S16 is executed.

[0127] S15, perform error counting.

[0128] The error counter is incremented by 1, and step 17 is executed.

[0129] errNum += 1

[0130] S16, perform correct counting.

[0131] The correct counter is incremented by 1.

[0132] correctNum += 1

[0133] S17, reset action counting related parameters.

[0134] The action counting related parameters are reset, and the next counting is prepared, and step S3 is executed.

[0135] fiveLeft = 0

[0136] sixRight = 0

[0137] toruchGround = 0

[0138] S18, detect whether the hands are on the right side of the second marker line 6.

[0139] Determine whether the hands are on the right side of the second marker line 6: substitute the 2D coordinates of leftHand and rightHand into equation two, if the results are both greater than 0, then the hands are on the right side of the second marker line 6, and step S19 is executed, otherwise, step S3 is executed.

[0140] S19, update the counter of the hands on the right side of the second marker line 6.

[0141] When it is detected that the hands are on the right side of the second marker line 6, the counter of the hands on the right side of the second marker line 6 is incremented by 1, and step S3 is executed.

[0142] S20, test end.

[0143] In this step, when the tester leaves is detected, indicating the end of the test, the correct number correctNum and the error number errNum are displayed on the interface, and after 10 seconds, the following data are reset, and then step S3 is executed.

[0144] Working state = detection

[0145] fiveLeft = 0; sixRight = 0;

[0146] touchGround = 0;

[0147] correctNum = 0; errNum = 0;

[0148] The number of actions is calculated based on the 2D coordinates of the human body key points, and whether a foul is detected, and the entire test process is executed in real time, and the processing speed is higher than 25 frames / second, real-time feedback is provided to the tester, so that the user can understand the test state in real time, and improper actions are corrected in time to achieve the best test result.

[0149] The beneficial effects of the present application are:

[0150] The hemispherical push-up arm row measuring method and implementation device calculate the number of actions based on the 2D coordinates of the human body key points, detect whether a foul is committed, the entire test process is executed in real time, and the processing speed is higher than 25 frames / second, real-time feedback is provided to the tester, so that the user can understand the test state in real time, and improper actions are corrected in time to achieve the best test result.

[0151] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0152] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A device for performing a semisphere roll and reach measurement, comprising: Including a column (1) and a test area (7), the bottom end of the column (1) is fixedly installed with a column base (2), and the side surface of the top end of the column (1) is fixedly installed with a smart screen (3); The test area (7) is provided with a hemisphere (4), and the side of the hemisphere (4) is respectively provided with a first mark line (5) and a second mark line (6) on the test area (7); The hemisphere prone push-up arm measurement method is operated based on the implementation device, and includes the following steps: S1, calibrate the mark line; S2, start testing: configure and start the camera, configure and start the human key point detection model, and then initialize the state variable; S3, receive a frame of image: read a frame of 1920x1080 MJPEG image from the camera, and decompress it into YUV420 format image, and finally send it to the human key point recognition model; S4, human key point recognition is performed; S5, judge the working state: if the current working state is detection, execute step S6, if the current working state is counting, execute step S8; S6, detect whether the tester is ready for testing; S7, set the working state to counting state, execute step S3; S8, detect whether the tester has left: if there is no human body frame or the human body frame does not meet the requirements, it means that the tester has left, then execute step S20, otherwise execute step S9; S9, detect whether the body touches the ground: if the body is detected to touch the ground, execute step S10, otherwise execute step S11; S10, update the touch ground counter; S11, detect whether the hand is on the left side of the first mark line (5); S12, update the counter of the hand on the left side of the first mark line (5); S13, judge whether the hand has passed the right side of the second mark line (6); S14, judge whether the body has touched the ground; S15, count the error; S16, count the correct; S17, reset the action counting related parameters; S18, detect whether the hand is on the right side of the second mark line (6); S19, update the counter of the hand on the right side of the second mark line (6); S20, test is over.

2. A device for performing the semisphere roll and arm stroke measurement according to claim 1, characterized in that, The installation position of the smart screen (3) faces the test area (7), the long side of the test area (7) is parallel to the long side of the smart screen (3), the smart screen (3) is built-in with a camera, and the camera is built-in with a human key point recognition model.

3. The apparatus of claim 1, wherein, The first mark line (5) and the second mark line (6) are arranged in parallel to each other, and both the mark lines are perpendicular to the long side of the test area (7), and the distance between the first mark line (5) and the second mark line (6) is 66 cm.

4. The apparatus of claim 1, wherein, In step S1, the calibration of the mark line includes: After the implementation device of the test is installed, the smart screen (3) displays the image captured by the camera in real time during calibration, and there are four movable points on the screen, and the four points are moved to the four endpoints of the two mark lines respectively, and then the determination button is clicked. After clicking OK, the 2D coordinates of the four endpoints are obtained, the two points with smaller x coordinates (x1, y1) and (x2, y2) determine the first mark line (5), and the two points with larger x coordinates (x3, y3) and (x4, y4) determine the second mark line (6), then the straight line equation corresponding to the first mark line (5) is equation one, and the straight line equation corresponding to the second mark line (6) is equation two; Equation one and equation two are as follows: (y2-y1)x+(x1-x2)y+(x2y1-x1y2)=0 equation one; (y4-y3)x+(x3-x4)y+(x4y3-x3y4)=0 equation two.

5. The apparatus of claim 1, wherein, In the step S4, the pre-processing of the human key point recognition model first down-samples the YUV420 image of 1920x1080 into an image of 960x540, and then feeds the image to the human key point recognition model. After the feature extraction, key point detection and post-processing of the human key point recognition model, the 19 2D key point coordinate positions and the confidence (x i , y i , c i ) of the human body are outputted, (x i , y i ) is the coordinate of the i-th key point, (c i ) is the confidence of the i-th key point, and the human body frame BBox(x, y, w, h, c) is outputted, (x, y) is the left upper corner coordinate of the human body frame, (w, h) is the width and height of the human body frame, and (c) is the confidence of the human body frame.

6. The apparatus of claim 1, wherein, In the step S6, the test process includes: First step: find the tester, since the test action is prone, the rule for finding the tester is: BBox.w>BBox.h and BBox.w>min_width and BBox.c>confidence_thres That is, the width of the human body frame is greater than the height of the human body frame, and the width of the human body frame is greater than the parameter min_width, and the confidence of the human body frame is greater than the parameter confidence_thres; If multiple human body frames satisfying the above rule are detected, the human body frame with the maximum value of (BBox.y+BBox.h), that is, the human body frame closest to the smart screen (3), is taken; Second step: detect whether the tester is ready, and the judgment condition includes that the hand is on the left side of the first mark line (5) and the body does not touch the ground.

7. The apparatus of claim 1, wherein, In the step S8, the human body frame or the human body frame does not satisfy the requirements as follows: BBox.w>BBox.h and BBox.w>min_width.

8. The apparatus of claim 1, wherein, In the step S10, when updating the touch ground counter, if the body is detected to touch the ground, the value of touchGroud is increased by 1, as shown in the following formula, and then step S11 is executed: touchGround+=1.

9. The apparatus of claim 1, wherein, In the step S20, when the tester is detected to leave, it indicates that the test is over, the correct number correctNum and the error number errNum are displayed on the interface, and after 10 seconds of display, the following data are reset, and then step S3 is executed: Working state=detection fiveLeft=0; sixRight=0; touchGround=0; correctNum=0; errNun=0.

Images