A vision-based method for detecting multiple sit-ups in a hallway

Abstract

The application discloses a kind of corridor in multi-person sit-up detection method based on vision, which is installed two cameras in the corridor between the same side wall interval, covers 12 test points, uses human key point detection technology to obtain tester head, crotch, knee, foot and other key point coordinates;Through calculating hook foot score, knee bending score, getting up score and lying down score four evaluation indexes, the action norm of tester is judged;System first extracts getting up interval, then searches lying down frame between adjacent getting up interval to form counting interval, and the proportion of hook foot and knee bending standard frame is counted in each counting interval, and when the proportion exceeds the set threshold value, it is determined as an effective sit-up;Finally, the number of sit-ups completed by the tester is obtained by accumulating all effective counting intervals.The application can realize simultaneous detection of multiple people, automatically judge the action norm, improve the detection efficiency and accuracy, and is suitable for sports testing, fitness training and other scenes.

Description

Technical Field

[0001] This invention relates to the field of physical fitness testing technology, specifically to a vision-based method for detecting sit-ups in multiple people within a corridor. Background Technology

[0002] With the development of physical education and the popularization of intelligent technology, sit-ups, as an important physical fitness test item, are constantly being improved in their detection methods. Traditional manual counting methods suffer from problems such as strong subjectivity and low efficiency. Therefore, intelligent detection technology based on computer vision has gradually become a research hotspot.

[0003] Existing vision-based sit-up detection technologies primarily employ single-camera or multi-camera systems for motion recognition and counting. For example, Chinese patent CN111401260B discloses a sit-up counting method based on the Quick-OpenPose model, which identifies key body parts by placing a camera on the side of the test area and uses the angle between the spine and the ground as a judgment criterion for counting. Chinese patent CN111368810B discloses a sit-up detection system based on the recognition of key points in the human body and skeleton, which determines the test subject's motion state by detecting changes in the angle between the straight line formed by key points in the waist and shoulders and the horizontal position. Chinese patent CN120022576A proposes a computer vision-based dual-camera sit-up testing method that uses two cameras for parallel detection and counting, enabling accurate counting and violation detection even without supervision. Furthermore, Chinese patents CN110732119A and CN115116126A also disclose methods and devices for automated sit-up testing using machine vision technology.

[0004] However, existing technologies have the following shortcomings: First, current intelligent sit-up detection systems are mainly designed for open spaces, typically requiring test subjects to be arranged horizontally in spacious stadiums or sports fields, with sufficient safety distance and camera field of view between each test position. This is difficult to achieve in primary and secondary schools where space resources are limited. Second, corridors, as common indoor spaces in schools, while long, are narrow. Existing technologies cannot effectively utilize such narrow spaces for simultaneous testing of multiple people. If individual testing is used, the detection efficiency will be low, making it difficult to meet the needs of centralized testing for classes or grades. Third, existing technologies lack optimized design for corridor environments in terms of judging the standardization of movements, especially in the comprehensive evaluation of key movement elements such as flexing the feet, bending the knees, standing up, and lying down. Therefore, there is an urgent need for a multi-person sit-up detection method suitable for narrow spaces such as corridors to improve detection efficiency and ensure detection accuracy. Summary of the Invention

[0005] To address the challenges of existing intelligent sit-up detection systems designed primarily for open spaces, which are often insufficient in primary and secondary schools with limited space, and the narrow width of corridors that prevent the simultaneous horizontal deployment of multiple test positions, as is the traditional method, and the inefficiency of individual testing by a single person, a vision-based method for multi-person sit-up detection in corridors is proposed. This method aims to enable simultaneous detection of multiple individuals within a narrow corridor, significantly improving efficiency and overcoming the reliance on open spaces inherent in traditional systems.

[0006] The objective of this invention is achieved through the following technical solution: a vision-based method for detecting sit-ups in multiple people within a corridor, comprising the following steps: S1. On the same side wall along the longitudinal direction of the corridor, two cameras are fixedly installed at 5m intervals. The angle between the optical axis of the camera and the ground, as well as the focal length, are set so that the horizontal field of view of a single camera is ≥4.8m and the fields of view of the two cameras do not overlap. Twelve test mats are laid vertically side by side along the longitudinal direction of the corridor. Each test mat corresponds to one test point and is numbered 1 to 12. Positions 1 to 6 fall into the field of view of camera a, and positions 7 to 12 fall into the field of view of camera b. Hook posts are fixed at each point on the side closest to the wall, 0.2m away from the wall, and the test subject's feet are facing the corresponding camera. S2. Before the test begins, facial images of each test subject are captured by a camera to complete identity recognition, and the corresponding height h is obtained by calling the pre-stored data. At the same time, the head coordinates (x3, y3), left hip coordinates (ulx0, uly0), right hip coordinates (urx0, ury0), and left and right end coordinates (x1, y1) and (x2, y1) of the hooked foot post are collected in the standard supine position. S3. During the detection process, the camera continuously acquires the coordinates of key human body points at various locations in a video frame sequence. These key points include the head coordinates (hx...). i hy i ), left hip coordinates (ulx) i uly i ), right hip coordinates (urx) i ,ury i Left knee coordinates (klx) i ,kly i ), right knee coordinates (krx) i ,kry i ), left foot coordinates (flx) i fly i ), right foot coordinate (frx) i fry i ); S4. For each frame i, calculate sequentially: Hooking foot score g1i When the x-coordinates of both the left and right feet are between x1 and x2 and the vertical pixel distance between the ankle and the hook post is less than the first threshold ts1, it means that the tester's feet are hooked on the post in accordance with the standard; otherwise, it is judged as non-standard. Knee flexion score g2 i When the ratio of the unilateral foot-hip distance to the leg length is less than the second threshold ts2, it indicates that the test subject's knee flexion is in accordance with the standard; otherwise, it is judged as non-standard. Get up and score in G3 i When the ratio of the relative displacement of the head and hips in the y direction to the length of the torso is less than the third threshold ts3, it indicates that the test subject's standing up is in accordance with the standard; otherwise, it is judged as non-standard. Lying down to score in G4 i When the head offset in the y-direction is less than the fourth threshold ts4 and the change in torso length is less than the fifth threshold ts5, it indicates that the test subject's lying position is in accordance with the standard; otherwise, it is judged as non-standard. S5, according to g3 i The score conforms to the specification of the starting frame i1 j With end frame i2 j Extract n3 starting intervals [i1] j i2 j [And search for the first time that g4 is satisfied between adjacent starting intervals and outside the first and last intervals.] i Frames whose scores conform to the specifications are used to form counting intervals. j ie j 】; S6. Within each counting interval, calculate the scores for the proportion of standard frames for hooking the foot and the proportion of standard frames for bending the knee, respectively. (g5) j When both are greater than the corresponding sixth threshold ts6 and seventh threshold ts7, it is determined that a valid sit-up has been completed in this interval. S7. Accumulate the number of all valid counting intervals to obtain the final number of sit-ups N completed by the test subject.

[0007] Preferably, in step S1, the camera is installed at a height of 2.0 m to 2.5 m above the ground, and the two cameras share a display screen installed in the middle of them for real-time feedback on the detection progress and results.

[0008] Preferably, in step S1, the length of the test pad is 1.8 m to 2.0 m, the width is 0.6 m to 0.8 m, and the center-line distance between adjacent test pads is 0.8 m to 1.0 m, so as to ensure that the longitudinal space utilization rate is ≥90%.

[0009] Preferably, in step S4, the hook foot score is g1. i Used to determine whether the test subject's feet are hooked on the post, when g1 i=1 indicates that hooking both feet onto the post conforms to the standard; g1 i It is calculated using the following formula: ; Wherein, ts1 is the set first judgment threshold, which is calculated by collecting images of the test subject's flexed feet based on the maximum vertical pixel distance from the ankle to the flexed foot post.

[0010] Preferably, in step S4, the knee flexion score g2 i Used to determine whether the test subject bends their knees, when g2 i =1 indicates that the test subject's knee flexion meets the standard; g2 i It is calculated using the following formula: ; Among them, ts2 is the set second judgment threshold, which is calculated by collecting the maximum value of the ratio of the distance between the foot and hip to the leg length when the test subject is normally bent at the knee; n1 is the scale between the physical distance jump and the pixel distance, that is, 1 meter corresponds to n1 pixel distance in the camera image.

[0011] Preferably, in step S4, the score for getting up is g3. i Used to determine whether the test subject gets up, when g3 i =1 indicates that the test subject's standing posture conforms to the standard; g3 i It is calculated using the following formula: ; ts3 is the set third judgment threshold, which is calculated by collecting the maximum value of the ratio of the distance from the test subject's head projection on the test pad to the hip bone and the distance to the torso when the test subject stands up.

[0012] Preferably, in step S4, the score g4 is obtained by lying down. i Used to determine whether the test subject is lying down, when g4 i =1 indicates that the test subject's lying position conforms to the standard; g4 i It is calculated using the following formula: ; Among them, ts4 is the fourth judgment threshold, which is calculated by collecting the maximum value of the vertical offset pixel distance of the head when the test subject lies down after completing the action under the standard action; ts5 is the fifth judgment threshold, which is calculated by collecting the maximum value of the torso length pixel distance difference when the test subject lies down after completing the action under the standard action.

[0013] Preferably, in step S5, .

[0014] Preferably, in step S6, the action quality score g5 in each counting interval j It is calculated using the following formula: ; Among them, ts6 is the set sixth judgment threshold, which is calculated by collecting the minimum value of the ratio of the number of frames that conform to the flexion of the feet to the total number of frames during the test subject's completion of a standard full sit-up; ts7 is the set seventh judgment threshold, which is calculated by collecting the minimum value of the ratio of the number of frames that conform to the knee flexion of the knee to the total number of frames during the test subject's completion of a standard full sit-up.

[0015] Preferably, in step S7, the number N of sit-ups completed by the test subject is calculated using the following formula: .

[0016] The beneficial effects of this invention are as follows: By optimizing the spatial layout of test subjects, the installation position and field of view of the cameras, and the corresponding multi-target recognition and action judgment algorithms, it fully utilizes the longitudinal space of the corridor to achieve simultaneous detection of multiple people. This provides schools with limited space with an efficient and convenient intelligent sit-up detection solution, promoting the widespread adoption of intelligent physical education testing. Compared with existing technologies, this invention can simultaneously detect 12 people in a narrow corridor space, greatly improving detection efficiency and solving the problem of traditional detection systems' dependence on open spaces. It is particularly suitable for the indoor physical education testing needs of primary and secondary schools during rainy, snowy, or low-temperature winter weather. Attached Figure Description

[0017] Figure 1 This is a top view of the test site used in the test method of this invention; Figure 2 This is a side view of the test site used in the test method of the present invention.

[0018] The attached diagram is labeled as follows: 1-Camera a, 2-Camera b, 3-Test pad, 4-Hook post, 5-Wall, 6-Tester, 7-Display screen. Detailed Implementation

[0019] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0020] See Figure 1-2 A vision-based method for detecting multiple people performing sit-ups in a corridor is proposed, which uses visual technology to automatically detect and count multiple people performing sit-ups simultaneously.

[0021] S1: Equipment Installation and Site Layout: Two cameras are fixedly installed at 5m intervals on the same side wall 5 along the longitudinal extension of the corridor. Cameras a1 and b2 are installed at a height of 2.0m to 2.5m above the ground. The angle between the camera's optical axis and the ground, as well as the focal length, are set to ensure that the horizontal field of view coverage of a single camera is ≥4.8m and that the fields of view of the two cameras do not overlap. Twelve test mats 3 are laid vertically side by side along the longitudinal direction of the corridor. The length of test mat 3 is 1.8m to 2.0m, and the width is 0.6m to 0.8m. The center-line spacing between adjacent test mats 3 is 0.8m to 1.0m to ensure that the longitudinal space utilization rate is ≥90%. Each test mat 3 corresponds to one test point and is numbered from 1 to 12. Positions 1 to 6 fall within the field of view of camera a1, and positions 7 to 12 fall within the field of view of camera b2. At each test point, a hook post 4 is fixed 0.2m away from the wall on the side closest to the wall, with the feet of the test subject 6 facing the corresponding camera. Two cameras share a display screen 7 installed between them for real-time feedback on the testing progress and results. This arrangement ensures that each camera can clearly capture human movements at the six test points, while avoiding interference caused by overlapping fields of view.

[0022] S2: Identity Recognition and Initial Data Acquisition: Before the detection begins, facial images of each of the 6 test subjects are captured via camera to complete identity recognition. Pre-stored data is then used to obtain the corresponding height h. Simultaneously, the coordinates of the head (x3, y3), left hip (ulx0, uly0), right hip (urx0, ury0), and the left and right ends (x1, y1) and (x2, y1) of the hook-foot post 4 are collected in a standard supine position. This step establishes personalized baseline data for subsequent motion detection, ensuring that the detection algorithm can accurately judge based on the body characteristics of each of the 6 test subjects.

[0023] S3: Real-time Keypoint Detection: During the detection process, the camera continuously acquires the coordinates of key human body points at various locations in the form of video frame sequences. Keypoints include head coordinates (hx). i hy i ), left hip coordinates (ulx) i uly i ), right hip coordinates (urx) i ,ury i Left knee coordinates (klx) i ,kly i ), right knee coordinates (krx) i ,kry i ), left foot coordinates (flx) i fly i ), right foot coordinate (frx) i fry iBy tracking the positional changes of these key points in real time, the system can accurately analyze the movement trajectory and posture changes of the test subject 6.

[0024] S4: Calculation of Action Standardization Score: For each frame i, four scoring indicators are calculated sequentially. Hooking foot score g1 i Used to determine whether the test subject's feet are hooked on the post, when g1 i =1 indicates that hooking both feet onto the post conforms to the standard, as shown by the formula: , The calculation yielded ts1, which is the set first judgment threshold, calculated by collecting images of the test subject's 6 flexed feet based on the maximum vertical pixel distance from the ankle to the flexion post 4. The knee flexion score g2... i Used to determine whether test subject 6 has bent their knees, when g2 i =1 indicates that test subject 6's knee flexion degree conforms to the standard, according to the formula: , The calculation yields ts2, which is the set second judgment threshold, calculated by collecting the maximum value of the ratio of the distance between the foot and hip to the leg length when the test subject 6 is normally knee-bent. n1 is the scale between physical distance and pixel distance, i.e., 1 meter corresponds to n1 pixels in the camera image. The standing score is g3. i Used to determine whether test subject 6 gets up, when g3 i =1 indicates that test subject 6's standing posture meets the standard, according to the formula: , The calculation yielded ts3, which is the set third judgment threshold, calculated by collecting the maximum value of the ratio of the distance from the test subject's head projection at position 3 on the test pad to the distance from the hip bone to the torso when the test subject stood up. The lying-down score is g4. i Used to determine whether test subject 6 is lying down, when g4 i =1 indicates that test subject 6's lying posture meets the standard, according to the formula: , The calculations show that ts4 is the fourth judgment threshold, which is calculated by collecting the maximum value of the vertical offset pixel distance of the head when lying down after completing the action of the test subject 6 standard actions, and ts5 is the fifth judgment threshold, which is calculated by collecting the maximum value of the difference in torso length pixel distance when lying down after completing the action of the test subject 6 standard actions.

[0025] S5: Getting-up interval recognition and counting interval division: based on g3 i The score conforms to the specification of the starting frame i1 j With end frame i2 j Extract n3 starting intervals [i1]j i2 j [And search for the first time that g4 is satisfied between adjacent starting intervals and outside the first and last intervals.] i Frames whose scores conform to the specifications are used to form counting intervals. j ie j 】

[0026] Specifically, first obtain the set of starting intervals, starting from i=1 and ending at the first interval that satisfies g3. i =1 in time frame i11, then from i11 until the first time g3 is satisfied. i =0 time frame i21+1, thus obtaining the first start interval [i11, i21], and so on to obtain each start interval [i11, i21]. j i2 j When there is a start time frame i1 j However, the participants remained standing until the end of the assessment, i.e., when i=n2 and When, then the last starting interval [i1] j i2 j =n2】, where j represents the sequence number of the stand-up interval, n2 is the time frame when the detection ends, n2 is the time frame when the detection ends, and the number of stand-up intervals is counted as n3; Then, between each starting interval j and j+1, where j=1,...,n3-1, from i2 j +1 to i1 j+1 The first one between -1 satisfies g4 i The time frame with value 1 is denoted as i3. j , reverse from i1 j+1 -1 to i2 j The first one between +1 satisfies g4 i =1 time frame is denoted as i4 j+1 Then, reverse the process from i11 to 1, starting with the first value that satisfies g4. i =1 time frame is recorded as i41; next, obtain the counting interval [is] corresponding to each starting interval j. j ie j The division of the counting interval is based on the formula: ; This interval division method ensures that each complete sit-up movement is accurately identified and segmented, providing precise time boundaries for subsequent movement quality assessment.

[0027] S6: Movement Quality Assessment: Within each counting interval, calculate the scores for the proportion of correct ankle flexion frames and the proportion of correct knee flexion frames, respectively (g5). j Through the formula: , The calculations yielded the following thresholds: ts6 is the sixth threshold, calculated as the minimum ratio of the number of frames conforming to the flexed-foot rule to the total number of frames during a standard sit-up performed by test subject 6; ts7 is the seventh threshold, calculated as the minimum ratio of the number of frames conforming to the knee flexion rule to the total number of frames during a standard sit-up performed by test subject 6. A valid sit-up is considered to have been completed within that interval when both thresholds ts6 and ts7 are greater than their respective values.

[0028] S7: Final Count Statistics: Accumulate the number of all valid count intervals to obtain the final number of sit-ups N completed by the test subject 6, using the formula... Calculated.

[0029] This testing method ensures the accuracy and fairness of sit-up counting through multi-dimensional assessment of movement standardization. The strategically placed cameras provide full coverage monitoring of all 12 testing points, while the key point detection algorithm accurately identifies the quality of each movement, effectively avoiding miscounting due to improper movements and improving the scientific rigor and reliability of the physical fitness test.

[0030] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A vision-based method for detecting sit-ups in multiple people within a corridor, characterized in that: Includes the following steps: S1. On the same side wall along the longitudinal direction of the corridor, two cameras are fixedly installed at 5m intervals. The angle between the optical axis of the camera and the ground, as well as the focal length, are set so that the horizontal field of view of a single camera is ≥4.8m and the fields of view of the two cameras do not overlap. Twelve test mats are laid vertically side by side along the longitudinal direction of the corridor. Each test mat corresponds to one test point and is numbered 1 to 12. Positions 1 to 6 fall into the field of view of camera a, and positions 7 to 12 fall into the field of view of camera b. Hook posts are fixed at each point on the side closest to the wall, 0.2m away from the wall, and the test subject's feet are facing the corresponding camera. S2. Before the test begins, facial images of each test subject are captured by a camera to complete identity recognition, and the corresponding height h is obtained by calling the pre-stored data. At the same time, the head coordinates (x3, y3), left hip coordinates (ulx0, uly0), right hip coordinates (urx0, ury0), and left and right end coordinates (x1, y1) and (x2, y1) of the hooked foot post are collected in the standard supine position. S3. During the detection process, the camera continuously acquires the coordinates of key human body points at various locations in a video frame sequence. These key points include the head coordinates (hx...). i hy i ), left hip coordinates (ulx) i uly i ), right hip coordinates (urx) i ,ury i Left knee coordinates (klx) i ,kly i ), right knee coordinates (krx) i ,kry i ), left foot coordinates (flx) i fly i ), right foot coordinate (frx) i fry i ); S4. For each frame i, calculate sequentially: Hooking foot score g1 i When the x-coordinates of both the left and right feet are between x1 and x2 and the vertical pixel distance between the ankle and the hook post is less than the first threshold ts1, it means that the tester's feet are hooked on the post in accordance with the standard; otherwise, it is judged as non-standard. Knee flexion score g2 i When the ratio of the unilateral foot-hip distance to the leg length is less than the second threshold ts2, it indicates that the test subject's knee flexion is in accordance with the standard; otherwise, it is judged as non-standard. Get up and score in G3 i When the ratio of the relative displacement of the head and hips in the y direction to the length of the torso is less than the third threshold ts3, it indicates that the test subject's standing up is in accordance with the standard; otherwise, it is judged as non-standard. Lying down to score g4 i When the head offset in the y-direction is less than the fourth threshold ts4 and the change in torso length is less than the fifth threshold ts5, it indicates that the test subject's lying position is in accordance with the standard; otherwise, it is judged as non-standard. S5, according to g3 i The score conforms to the specification of the starting frame i1 j With end frame i2 j Extract n3 starting intervals [i1] j i2 j [And search for the first time that g4 is satisfied between adjacent starting intervals and outside the first and last intervals.] i Frames whose scores conform to the specifications are used to form counting intervals. j ie j 】; S6. Within each counting interval, calculate the scores for the proportion of standard frames for hooking the foot and the proportion of standard frames for bending the knee, respectively. (g5) j When both are greater than the corresponding sixth threshold ts6 and seventh threshold ts7, it is determined that a valid sit-up has been completed in this interval. S7. Accumulate the number of all valid counting intervals to obtain the final number of sit-ups N completed by the test subject.

2. The vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, is characterized in that: In step S1, the camera is installed at a height of 2.0 m to 2.5 m above the ground, and the two cameras share a display screen installed in the middle of them for real-time feedback on the detection progress and results.

3. The vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, is characterized in that: In step S1, the length of the test pad is 1.8 m to 2.0 m, the width is 0.6 m to 0.8 m, and the center-line distance between adjacent test pads is 0.8 m to 1.0 m to ensure that the longitudinal space utilization rate is ≥90%.

4. The vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, is characterized in that: In step S4, the score for hooking the foot is g1. i Used to determine whether the test subject's feet are hooked on the post, when g1 i =1 indicates that hooking both feet onto the post conforms to the standard; g1 i It is calculated using the following formula: ； Wherein, ts1 is the set first judgment threshold, which is calculated by collecting images of the test subject's flexed feet based on the maximum vertical pixel distance from the ankle to the flexed foot post.

5. The vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, is characterized in that: In step S4, the knee flexion score is g2. i Used to determine whether the test subject bends their knees, when g2 i =1 indicates that the test subject's knee flexion meets the standard; g2 i It is calculated using the following formula: ； Among them, ts2 is the set second judgment threshold, which is calculated by collecting the maximum value of the ratio of the distance between the foot and hip to the leg length when the test subject is normally bent at the knee; n1 is the scale between the physical distance jump and the pixel distance, that is, 1 meter corresponds to n1 pixel distance in the camera image.

6. The vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, is characterized in that: In step S4, the score for getting up is g3. i Used to determine whether the test subject gets up, when g3 i =1 indicates that the test subject's standing posture conforms to the standard; g3 i It is calculated using the following formula: ； ts3 is the set third judgment threshold, which is calculated by collecting the maximum value of the ratio of the distance from the test subject's head projection on the test pad to the hip bone and the distance to the torso when the test subject stands up.

7. The vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, is characterized in that: In step S4, lying down scores g4. i Used to determine whether the test subject is lying down, when g4 i =1 indicates that the test subject's lying position conforms to the standard; g4 i It is calculated using the following formula: ； Among them, ts4 is the fourth judgment threshold, which is calculated by collecting the maximum value of the vertical offset pixel distance of the head when the test subject lies down after completing the action under the standard action; ts5 is the fifth judgment threshold, which is calculated by collecting the maximum value of the torso length pixel distance difference when the test subject lies down after completing the action under the standard action.

8. A vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, characterized in that: In step S5, 。 9. A vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, characterized in that: In step S6, the motion quality score g5 in each counting interval is calculated. j It is calculated using the following formula: ； Among them, ts6 is the set sixth judgment threshold, which is calculated by collecting the minimum value of the ratio of the number of frames that conform to the flexion of the feet to the total number of frames during the test subject's completion of a standard full sit-up; ts7 is the set seventh judgment threshold, which is calculated by collecting the minimum value of the ratio of the number of frames that conform to the knee flexion of the knee to the total number of frames during the test subject's completion of a standard full sit-up.

10. A vision-based method for detecting sit-ups in a corridor using multiple people, as described in claim 1, characterized in that: In step S7, the number of sit-ups N completed by the test subject is calculated using the following formula: 。

Images