CN114307117B - Standing long jump result measuring method and device based on video - Google Patents
Standing long jump result measuring method and device based on video Download PDFInfo
- Publication number
- CN114307117B CN114307117B CN202111632521.5A CN202111632521A CN114307117B CN 114307117 B CN114307117 B CN 114307117B CN 202111632521 A CN202111632521 A CN 202111632521A CN 114307117 B CN114307117 B CN 114307117B
- Authority
- CN
- China
- Prior art keywords
- athlete
- long jump
- posture
- landing
- video
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a technical method and a device for measuring standing long jump performance based on video. The method designs a standing long jump performance measuring algorithm based on video, extracts the human body posture information of the athlete in the video through human body posture detection, and judges the take-off and landing actions of the athlete according to the human body posture change; by optimizing the tracking mode of the human body posture, the problem of tracking number exchange under the background of multiple persons is solved; the rear edge of the landing point of the athlete is accurately positioned by combining various image information such as frame difference, edge, frequency domain and the like; and accurately calculating the standing long jump performance of the athlete by combining the parameters such as the side length of the long jump area and the like through inverse perspective transformation. The long jump process of athletes with different heights and long jump actions is shot by a high-resolution camera, and the average absolute error of experimental measurement is 0.69cm. The method and the device for measuring the standing long jump performance based on the video avoid using expensive infrared equipment and do not need complicated later maintenance.
Description
Technical Field
The invention belongs to the technical field of video image processing, and particularly relates to a standing long jump result measuring method and device based on a video.
Background
With the development of artificial intelligence and computer vision, artificial intelligence devices based on video images play an increasingly important role in human life. Some manual measurement links can use visual intelligent equipment to carry out measurement, thereby avoiding the consumption of labor cost and saving time cost.
The country attaches more importance to and supports sports, and the standardization of sports examinations also becomes the key point. If can directly confirm examinee's sports score through sports examination video, can use manpower sparingly and avoid the consumption of unnecessary equipment, the process of sports examination can be preserved through the video simultaneously, and the later stage is examined validity and the accuracy of sports examination score at any time, has avoided the emergence of some cheating behaviors, also conveniently deals with some questions of examinee. Standing long jump is a common sports project and a common sports examination project, and is important for intellectualization of measurement of the standing long jump examination result.
In the existing standing long jump test, a commonly used measuring device is an infrared sensing device. The price of the equipment is more than four thousand to ten thousand, and the later maintenance cost of the infrared equipment is very expensive. The price of a high-resolution camera is only two or three hundred, and the standing long jump performance can be measured only by carrying a notebook computer and connecting the camera.
Disclosure of Invention
The invention provides a video-based standing long jump performance measuring method and device, aiming at the problems that standing long jump performance measurement cannot be rechecked, the current measuring equipment is expensive and the like.
The invention is realized by adopting the following technical scheme:
the standing long jump result measuring method based on the video comprises the following steps:
s1, acquiring an image without shielding a long jump test area, and extracting the long jump test area;
s2, acquiring a standing long jump video in real time, detecting and tracking the posture of the athlete, and judging the standing long jump state of the athlete;
s3, accurately positioning the rear edge of the landing point of the athlete and measuring the long jump performance;
s4, measuring the actual distance from the positioning point to the jumping wire, namely the measured standing long jump score;
and S5, displaying the standing long jump state and the result measurement result of the athlete.
Preferably, in S1 and S2, the acquired video image is a high-resolution image, specifically an image with a resolution of 3264 × 2448.
Preferably, in S2, the multi-person posture detector OpenPose is used to detect the human body posture information in real time.
Preferably, in S2, after the multi-person posture information is detected, the posture ID is updated by using an improved posture tracking strategy, and the improved posture tracking strategy is divided into two stages:
when the posture ID of the athlete is not determined, the average distance between key points corresponding to each posture in the previous frame and the next frame is used as the similarity of the postures in the two frames, the postures with the maximum similarity are matched with each other, and the posture ID of the next frame is updated by using the posture ID of the previous frame in the matching pair; when the posture of a certain ID is continuously positioned in the standing long jump test zone, the ID is determined as the player posture ID; when the continuous multiframes are positioned outside the standing long jump test area, the athlete attitude ID is emptied;
when the athlete posture ID is determined, only the athlete posture ID is tracked; the similarity of the stage is obtained by calculating the IOU, the motion direction and the motion amplitude according to the detection result of the posture of the athlete in the previous frame and the detection result of each posture in the next frame; if the pose of the next frame is positioned in the standing long jump test area, adding extra similarity to the pose; and taking the posture with the highest similarity in the next frame and higher than the threshold value as the tracking result of the posture of the athlete.
Preferably, in S2, the posture detection key points are analyzed, that is, the standing long jump state of the athlete is determined according to the angle changes of the thigh and the calf and the displacement of the foot; only four key points of the trunk center, the right thigh root, the right knee joint and the right ankle of the posture of the athlete are used for analysis; if the angles of the thighs and the trunk and the angles of the thighs and the shanks are continuously reduced for 3 times, and the foot displacement exceeds a threshold value, the jumping is considered to be finished; after jumping, the foot displacement is smaller than the threshold value, and the landing action is considered to be finished.
Preferably, in S3, a criterion for continuously positioning the landing site is provided, that is, after the mobile completes the landing action, the rear edge of the landing site is continuously positioned until the positioning stop condition is met; two conditions for meeting the positioning and stopping of the landing point are provided, one is that the athlete stands stably, and the posture of the athlete shows that the angle of the thigh and the shank is larger than a threshold value and the foot displacement is smaller than the threshold value; the other is that the athlete leaves the landing point, and the posture of the athlete shows that the displacement of the posture central point is larger than the width of the posture detection frame when the athlete lands on the ground.
Preferably, in S3, an anti-occlusion criterion is proposed, and for a frame with occlusion, no landing point positioning is performed; the criterion condition of the shielding prevention is that if the distance from the thigh root key point to the ankle key point is smaller than a threshold value, the hip is considered to shield the foot, and the frame does not position the rear edge of the landing point.
Preferably, in S3, the specific implementation method for precisely positioning the rear edge of the landing site is as follows:
after the athlete finishes landing, firstly obtaining a motion segmentation map by a KNN background difference method, obtaining the position of the athlete in a picture by the position of a human body posture detection frame, determining an effective area of a landing point by setting a long jump test area, extracting the motion segmentation map of an intersection area of the athlete posture detection frame and the long jump test area, and carrying out morphological processing on the motion segmentation map to obtain a rough segmentation map only containing feet of the athlete, namely landing parts; carrying out inverse perspective transformation on the rough segmentation image and the color original image to obtain a vertical visual angle image of the long jump test area; extracting the foot boundary of the vertical view of the rough segmentation graph, extracting the vertical view of the color graph in the corresponding boundary, carrying out bilateral filtering on the vertical view to remove the influence of the ground texture, then carrying out secondary wavelet transformation, extracting horizontal high-frequency details to obtain a detail graph of the shoe, and carrying out canny edge detection on the detail graph to obtain the edge of the shoe; and calculating the intersection part of the vertical visual angle graph of the rough segmentation graph and the canny detection graph, wherein the point set which is closest to the left side, namely the jumping line is judged as the rear edge of the landing point of the athlete.
The standing long jump achievement measuring device based on the video comprises:
the video acquisition module is used for acquiring the standing long jump video frame in real time through the high-resolution camera and is used for subsequently judging the standing long jump process and the standing long jump performance of the athlete;
the test area extraction module is used for segmenting the standing long jump test area by threshold segmentation and reservation of a maximum connected domain;
the take-off and landing judgment module detects the posture of the athlete through the human body posture detector, updates the posture ID by using an improved posture tracking strategy, and analyzes the posture to obtain the motion state of the athlete, namely whether the take-off action is finished or not and whether the landing action is finished or not;
the landing point positioning module extracts a background difference image firstly and extracts high-frequency details by secondary wavelet transformation so as to position a rear edge point of a landing point of an athlete for standing long jump performance measurement;
the score measuring module is used for calculating standing long jump scores by applying inverse perspective transformation in combination with the known long jump area size and the extracted long jump test area;
and the result display module is used for displaying the standing long jump state of the athlete, including take-off, landing and finishing, and standing long jump scores at the upper right corner of the image.
The invention has at least the following beneficial technical effects:
the standing long jump result measuring device based on the video provided by the invention can accurately and quickly realize the measurement of the standing long jump result. Through actual multiple standing long jump tests of multiple persons, the accuracy of the measurement result reaches centimeter level, and the processing time of each standing long jump is about 10 seconds on average. In addition, common athletes with different heights and different long jump distances are tested under the conditions that long jump backgrounds are complex and contain interfering figure backgrounds, the accuracy of the measurement result of the standing long jump performance is high, the speed is high, good robustness is shown, and the requirement of the current intelligent sports test on the measurement of the standing long jump test performance is basically met.
Further, in the posture detection section, a multi-person posture detector openposition is used to detect the posture of the athlete in the video. In the standing long jump performance measuring method, key points of the legs of the athletes are mainly concerned. Because the camera carries out video acquisition from the athlete's right side, only the posture of the leg on the right side is selected for analysis. By adopting the openposition attitude detector capable of detecting the postures of multiple persons, the influence of background persons on the posture detection of the athletes is avoided, and the high human body posture detection accuracy is obtained under the condition of real-time performance.
Further, in the gesture tracking part, a gesture tracking strategy is improved, and tracking is carried out by using the improved gesture tracking strategy. The improved posture tracking strategy focuses on tracking the postures of the athletes, and after the postures of the athletes are numbered, only the numbered postures are tracked and matched. Before the athlete posture number is determined, the similarity of the postures in the two frames is directly calculated according to the distance between the corresponding key points of each posture in the two frames, and matching is carried out according to the similarity from high to low. After the athlete attitude number is determined, after each frame of image is subjected to attitude detection, the matching degree is calculated through an IOU (Intersection over Unit), the motion direction and the motion amplitude. In addition, if the gesture to be matched of the frame is located in the standing long jump test area, additional matching degree addition is given to the gesture. The posture tracking mode can greatly reduce the influence of background characters on the posture detection of the athlete, and even if the posture of the athlete is not detected in a certain frame or the detected posture of the athlete has a large error, the posture of the athlete which is correctly detected can be tracked in the next frame.
Further, in the take-off and landing judgment part, the standing long jump state of the athlete is judged by analyzing the posture detection key points, namely according to the angle changes of thighs and shanks and the displacement of feet. Only four key points of the trunk center, the right thigh root, the right knee joint and the right ankle of the posture of the athlete are used for analysis. If the angles of the thighs and the trunk and the angles of the thighs and the shanks are continuously reduced for 3 times, and the foot displacement exceeds a threshold value, the jumping is considered to be finished; after jumping, the foot displacement is smaller than the threshold value, and the landing action is considered to be finished.
Furthermore, a continuous positioning criterion is proposed in the landing point positioning part, namely whether to finish the landing point positioning is judged according to the posture and the motion amplitude of the athlete. Two judgment conditions are provided, wherein firstly, the angles of thighs and shanks of athletes are larger than a certain angle, and the foot displacement is smaller than a certain threshold value; secondly, the displacement of the athlete exceeds the maximum width of the posture at the moment when the athlete lands on the ground. If only one of the above two conditions is satisfied, the athlete is considered to have completed all the long jump actions, and therefore the positioning of the landing site is finished. This section ensures that the measurement of performance is continually updated when the athlete falls unstable.
Furthermore, a blocking removing criterion is provided at the landing point positioning part, whether blocking occurs is judged according to the posture of the athlete when landing, and the frame with blocking does not carry out landing point positioning. And taking the distance between the key points of the hip and the foot as a shielding criterion, and when the distance between the hip and the foot is smaller than a certain threshold value, considering that hip shielding occurs, and not positioning the landing point of the frame.
Furthermore, in the landing point positioning part, the background difference image, the frequency domain information, the edge information and the like are combined for processing, and the landing point of the moving person is accurately positioned. After finishing landing action of the athlete, firstly obtaining a motion segmentation map by a KNN background difference method, obtaining the position of the athlete in a picture by the position of a human body posture detection frame, determining an effective area of a landing point by setting a long jump test area, only extracting the motion segmentation map of the intersection area of the athlete posture detection frame and the long jump test area, and carrying out morphological processing on the motion segmentation map to obtain a rough segmentation map only containing the foot part (landing part) of the athlete. And carrying out inverse perspective transformation on the rough segmentation image and the color original image to obtain a vertical visual angle image of the long jump test area. The method comprises the steps of extracting foot boundaries of vertical view images of rough segmentation images, extracting vertical view images of color images in corresponding boundaries, carrying out bilateral filtering on the foot boundaries to remove influences of ground textures, then carrying out secondary wavelet transformation, extracting horizontal high-frequency details to obtain a detail image of the shoe, and carrying out canny edge detection on the detail image to obtain the edge of the shoe. And calculating the intersection part of the vertical visual angle graph of the rough segmentation graph and the canny detection graph, wherein the point set which is closest to the left side, namely the jumping line is judged as the rear edge of the landing point of the athlete. In the part, the rear edge of the landing point of the athlete is accurately determined, and a foundation is laid for accurately measuring the long jump performance subsequently.
Further, in the achievement measuring section, the standing long jump achievement is calculated by inverse perspective transformation. And 4 angular points of the test area are extracted from the test area extracted from the front part, inverse perspective transformation is carried out by combining the actual size of the test area, the actual distance from the positioning point to the edge of the long jump ground mat is obtained, the distance from the jumping line to the edge of the long jump ground mat is reduced, and the achievement of the athlete in the standing long jump is obtained. The precision of the result measurement result is higher, the precision of centimeter level is achieved, and the requirement of 1 centimeter precision in the standing long jump examination is basically met.
Further, in the achievement display section, the long jump status and the measurement achievement of the athlete are displayed in the upper right-hand upper part of the screen. The long jump status (take-off, landing, finish) of the athlete is displayed in the first row at the upper right of the picture, and long jump scores are displayed in the second row at the upper right of the picture. The standing long jump score can be visually displayed, and recording and verification are facilitated.
In conclusion, the video-based standing long jump performance measuring method and device disclosed by the invention avoid the use of expensive infrared equipment, do not need complicated later maintenance, can still realize high-precision measuring results and rapid processing speed for common athletes with different heights and different long jump habits and under the condition that people are interfered in a long jump background, and have the advantages that the final standing long jump performance measuring precision reaches the centimeter level, the average processing time is about 10 seconds, the video-based standing long jump performance measuring method and device meet the requirements of standing long jump performance measurement in sports examinations, and the video-based standing long jump performance measuring method and device have good application value in intelligent sports examination yards.
Drawings
Fig. 1 is a view showing a structure of a video-based standing long jump performance measuring apparatus according to the present invention.
Fig. 2 is a flow chart of a video-based standing long jump performance measuring method in the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art. It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
Example 1
As shown in fig. 1, the device for measuring standing long jump performance based on video provided by the invention comprises the following functions or modules: the device comprises a video acquisition module, a test area extraction module, a take-off and landing judgment module, a landing point positioning module, a score measurement module and a result display module.
The output end of the video acquisition module is connected with the input end of the test area extraction module, the input end of the take-off and landing judgment module, the input end of the landing point positioning module is connected with the input end of the result display module, the output end of the test area extraction module is connected with the input end of the take-off and landing judgment module, the input end of the landing point positioning module and the input end of the score measurement module, the output end of the take-off and landing judgment module is connected with the input end of the landing point positioning module, the output end of the landing point positioning module is connected with the input end of the score measurement module, and the output end of the score measurement module is connected with the input end of the result display module.
The video acquisition module is used for acquiring standing long jump video frames in real time through the high-resolution camera and is used for subsequently judging the standing long jump process and the standing long jump performance of the athlete;
the test area extraction module is used for segmenting the standing long jump test area by threshold segmentation and reservation of a maximum connected domain;
the take-off and landing judgment module detects the posture of the athlete through the human body posture detector, updates the posture ID by using an improved posture tracking strategy, and analyzes the posture to obtain the motion state of the athlete, namely whether the take-off action is finished or not and whether the landing action is finished or not;
the landing point positioning module extracts a background difference image firstly and extracts high-frequency details by secondary wavelet transformation so as to position a rear edge point of a landing point of an athlete and is used for standing long jump performance measurement;
the score measuring module is used for calculating standing long jump scores by applying inverse perspective transformation in combination with the known long jump area size and the extracted long jump test area;
and the result display module is used for displaying the standing long jump state of the athlete, including take-off, landing and finishing, and standing long jump scores at the upper right corner of the image.
Example 2
The video-based standing long jump result measuring method of the invention has a flow chart shown in fig. 2, and comprises video acquisition, test area extraction, attitude detection, attitude tracking, take-off and landing judgment, continuous positioning judgment, anti-blocking judgment, landing point positioning, result measurement and result display.
In the video acquisition part, in order to improve the accuracy of measuring the standing long jump performance, a camera with large resolution is selected for acquisition. In the experiment, a camera with the resolution of 3264 multiplied by 2448 is used, the frame rate is 15 frames per second, and the compression format is MJPEG. The installation height of the camera is 205cm, the corner point closest to the long jump test frame is taken as a reference, the long jump direction is taken as the positive x direction on the horizontal plane, and the direction vertical to the x direction is taken as the y direction. The horizontal x distance from the standing jump test area is 180cm, and the horizontal y distance is-5 cm. And starting to run the program, and starting to acquire images by the camera and display the images in the window. In order to show the most accurate measurement result, the angle of the camera is adjusted, and the standing long jump test area is close to the right edge and the lower edge of the picture of the camera as much as possible.
In the test area extraction part, after the position and the angle of the camera are fixed, the reset button is clicked to collect an image under the condition that the standing long jump test area is not shielded, and the long jump test area is divided through the color difference between the long jump pad and the ground. After the acquisition of the test area is finished, the athlete prepares for long jump before entering the test area to take off the jumping line, and at the moment, the athlete clicks a start button to start processing a long jump video frame. After the camera is moved or the long jump pad is moved each time, the reset button needs to be clicked, and the standing long jump test area is extracted again.
The gesture detection part adopts a human body gesture detector openposition which can quickly detect the gestures of the athletes in the video. The method uses the posture detector to judge whether the athlete takes off and lands in the standing long jump process, and is helpful for judging the long jump process, mainly the posture changes of the legs and arms of the athlete. For objects moving in a fast and large range, the ambiguity in the collected video frame is high, the detection accuracy of corresponding key points is low, and the arms of the athlete always move in a fast and large range in the standing long jump movement, so the method for measuring the standing long jump performance mainly focuses on the key points of the legs of the athlete. If carry out video acquisition from sportsman's right side, then sportsman's left side limbs are sheltered from, detect the precision and descend greatly, consequently only select right side shank gesture to carry out the analysis. The openposition posture detector capable of detecting the postures of multiple persons is adopted to avoid the influence of background persons on the posture detection of the athletes. In order to increase the processing speed of the gesture detection part, the collected 3264 × 2448 resolution video frame is firstly down-sampled to 1632 × 1224 resolution, and then gesture detection is performed.
In the gesture tracking part, similarity calculation is carried out on a plurality of gestures detected by two video frames before and after, and matching is carried out on the gestures with the highest similarity, so that the gesture tracking is completed. Because the method only focuses on the postures of the athletes, only the posture number of the athlete needs to be found, and the tracking matching is continuously carried out on the numbered postures. Because the background character can not be in the standing long jump test area, the posture numbers of the athletes are recorded according to the posture numbers, when the posture of a certain number enters the test area, the posture corresponding to the number is considered as the posture of the athlete, and then only the posture of the number is tracked. Before the athlete posture number is determined, the similarity of the postures in the two frames is directly calculated according to the distance between the corresponding key points of each posture in the two frames, and matching is carried out according to the similarity from high to low. After the player posture number is determined, after the posture detection is carried out on each frame of image, only the posture which is matched with the posture of the previous frame of player is searched for and used as the tracking of the player posture. The matching degree at this time is calculated by an IOU (Intersection over Unit), a motion direction and a motion amplitude, and in addition, if the posture to be matched of the frame is located in the standing long jump test area, extra matching degree addition is added to the posture. And if the matching degrees of all the postures are lower than the threshold value, determining that no posture matched with the player exists. And if the continuous multi-frame does not have the matched postures of the athletes, resetting the posture number of the athletes. The posture tracking mode can greatly reduce the influence of background characters on the posture detection of the athlete, and even if the posture of the athlete is not detected in a certain frame or the detected posture of the athlete has a large error, the posture of the athlete which is correctly detected can be tracked in the next frame.
And analyzing the takeoff and landing judging part according to the obtained attitude information. It has been observed that when an athlete starts to set a long jump, the angle of his thigh and lower leg is continuously decreased, and then if the foot displacement of the athlete exceeds a certain threshold, the athlete is considered to have finished taking off. After the jumping action is finished, when the foot displacement of the athlete is smaller than a certain threshold value, the athlete is considered to finish the landing action. And when the athlete finishes the landing action, the athlete landing point is extracted.
And in the continuous positioning judgment part, whether the landing point positioning is finished or not is judged according to the posture and the motion amplitude of the athlete. In a standing long jump test, a situation that an athlete backs one step after landing or sits on the ground often occurs, and the long jump score is measured according to the back landing point. Therefore, the landing point of the athlete at the moment of landing is not necessarily the final performance, and the measurement is continuously carried out for a period of time after the athlete lands until the athlete completely finishes the long jump. Because different athletes have different long jump habits, it is unreasonable to continuously position the athlete directly according to the duration, so that two judgment conditions are added, namely that the angles of thighs and shanks of the athletes are larger than a certain angle and the foot displacement is smaller than a certain threshold value; secondly, the maximum width of the posture of the athlete at the moment when the displacement exceeds the ground. If only one of the above two conditions is satisfied, the athlete is considered to have completed all the long jump actions, and therefore the positioning of the landing site is finished.
And the occlusion removal judgment part judges whether occlusion occurs or not according to the posture of the athlete when landing, and removes the long jump score measured under the occlusion. After some athletes fall to the ground, the athlete squats deeply to keep stability, the buttocks of the athletes do not touch the ground during the deep squat, but due to the fact that the camera is installed to be high, the part, closest to the jumping line, of the collected images is the buttocks of the athletes, and the final measuring result of the long jump result is smaller than the actual long jump result. In order to solve the problem, an occlusion criterion is provided, and the determined frame with occlusion does not carry out landing point positioning and result measurement. The gesture detection of the athlete comprises key points such as buttocks and feet, the distance between the critical points of the buttocks and the feet is used as a shielding criterion, when the distance between the buttocks and the feet is smaller than a certain threshold value, the buttocks shielding is considered to be generated, and the landing point positioning and the achievement measurement are not carried out in the frame.
And the landing point positioning part is combined with the background difference image, the frequency domain information, the edge information and the like for processing, so as to accurately position the landing point of the moving member. In order to improve the accuracy of the final standing long jump performance measurement, video frames with the original resolution, namely images with the resolution of 3264 × 2448, are used in the landing point positioning and the subsequent long jump performance measurement. After the athlete finishes the landing action, firstly obtaining a motion segmentation map by a KNN background difference method, obtaining the position of the athlete in a picture by the position of a human body posture detection frame, determining an effective area of a landing point by setting a long jump test area, only extracting the motion segmentation map of the intersection area of the athlete posture detection frame and the long jump test area, and carrying out morphological processing on the motion segmentation map to obtain a rough segmentation map only containing feet (landing parts) of the athlete. And carrying out inverse perspective transformation on the rough segmentation image and the color original image to obtain a vertical visual angle image of the long jump test area. The method comprises the steps of extracting foot boundaries of vertical view images of rough segmentation images, extracting vertical view images of color images in corresponding boundaries, carrying out bilateral filtering on the foot boundaries to remove influences of ground textures, then carrying out secondary wavelet transformation, extracting horizontal high-frequency details to obtain a detail image of the shoe, and carrying out canny edge detection on the detail image to obtain the edge of the shoe. And calculating the intersection part of the vertical visual angle graph of the rough segmentation graph and the canny detection graph, wherein the point set which is closest to the left side, namely the jumping line is judged as the rear edge of the landing point of the athlete.
The achievement measuring part calculates the achievement of standing long jump by inverse perspective transformation. The method is used on the premise that the size of the long jump test area, namely the length and the width of the long jump test area and the distance from the jumping line to the edge of the long jump test area are known. The part divides the long jump test area to obtain 4 angular points of the long jump test area, and the inverse perspective transformation is carried out by combining the actual size of the long jump ground mat to obtain an image vertical to the ground visual angle. The positioning point of the landing point is also converted to the image, the pixel-level distance from the positioning point to the edge of the image can be directly obtained, the actual distance from the positioning point to the edge of the long jump testing area is obtained by referring to the image pixel ruler and the actual size of the long jump testing area, the distance from the jumping line to the edge of the long jump testing area is reduced, and the achievement of the athlete in the current set long jump is obtained. In the process of continuously positioning the landing site, if a closer long jump distance is measured, the long jump performance of the body is updated.
In the achievement display part, the long jump state and the measurement achievement of the athlete are displayed on the upper right of the picture in an ascending and descending way. The athlete's long jump status will be displayed in the first row on the top right of the frame and long jump performance will be displayed in the second row on the top right of the frame. When the athlete finishes the take-off action, the long jump state 'take-off' displayed in the first row at the upper right of the picture; when the athlete finishes the landing action, the first line at the upper right of the picture displays the long jump state of landing; when the athlete finishes the long jump action, the first row at the upper right of the Chinese shows that the long jump state is 'finished'. When the long jump score is not measured, for example, the athlete has not taken off, the score of the second line at the upper right shows an initial value of 0 meter; after the long jump score of the athlete is measured, the measured long jump distance is displayed on the second line score at the upper right part, and if the measured long jump distance is reduced due to reasons such as backward movement of the athlete and the like, the score is immediately updated on the second line at the upper right corner of the picture; and after the first long jump state in the upper right corner displays 'finish', the long jump achievement in the second line is the final long jump achievement.
The speed of the method does not reach real-time performance in the whole process, but the final performance can be obtained within 10 seconds after the athlete finishes long jump under most conditions. The video acquisition, attitude detection, attitude tracking, take-off and landing judgment parts of the method can achieve real-time performance, and the landing point positioning part does not have real-time performance.
The present invention measures standing long jump performance of 12 different athletes, with the test times and results shown in table 1 below. The result shows that in 12 long jump tests, the error of 9 tests is less than 1 cm, the error of the other 3 tests is less than 2 cm, the average absolute error is 0.6979 cm, and the requirement that the accuracy of the achievement of a standing long jump test is 1 cm can be basically met.
TABLE 1 standing long jump result measurement results
Although the invention has been described in detail hereinabove with respect to a general description and specific embodiments thereof, it will be apparent to those skilled in the art that modifications or improvements may be made thereto based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.
Claims (3)
1. The standing long jump result measuring method based on the video is characterized by comprising the following steps of:
s1, acquiring an image without shielding a long jump test area, and extracting the long jump test area;
s2, acquiring a standing long jump video in real time, detecting and tracking the posture of the athlete, and judging the standing long jump state of the athlete; after multi-person posture information is detected, posture ID updating is carried out by using an improved posture tracking strategy, and the improved posture tracking strategy is divided into two stages:
when the posture ID of the athlete is not determined, the average distance between key points corresponding to each posture in the previous frame and the next frame is used as the similarity of the postures in the two frames, the postures with the maximum similarity are matched with each other, and the posture ID of the next frame is updated by using the posture ID of the previous frame in the matching pair; when the posture of a certain ID is continuously positioned in the standing long jump test zone, the ID is determined as the player posture ID; when the continuous multiframes are positioned outside the standing long jump test area, the athlete attitude ID is emptied;
when the athlete posture ID is determined, only the athlete posture ID is tracked; the similarity of the stage is obtained by calculating the IOU, the motion direction and the motion amplitude according to the detection result of the posture of the athlete in the previous frame and the detection result of each posture in the next frame; if the pose of the next frame is positioned in the standing long jump test area, adding extra similarity to the pose; taking the posture with the highest similarity and higher than the threshold value in the next frame as the tracking result of the posture of the athlete; the posture detection key points are analyzed, namely the standing long jump state of the athlete is judged according to the angle changes of thighs and shanks and the displacement of feet; only four key points of the trunk center, the right thigh root, the right knee joint and the right ankle of the posture of the athlete are used for analysis; if the angles of the thighs and the trunk and the angles of the thighs and the shanks are continuously reduced for 3 times, and the foot displacement exceeds a threshold value, the jumping is considered to be finished; after jumping, if the foot displacement is less than the threshold value, the landing action is considered to be finished;
s3, accurately positioning the rear edge of the landing point of the athlete and measuring the long jump performance; a criterion of continuously positioning the landing point is provided, namely after the sportsman finishes the landing action, the rear edge of the landing point is continuously positioned until the positioning stop condition is met; two conditions for meeting the positioning and stopping of the landing point are provided, one is that the athlete stands stably, and the posture of the athlete shows that the angle of the thigh and the shank is larger than a threshold value and the foot displacement is smaller than the threshold value; the other is that the athlete leaves the landing point, and the posture of the athlete is represented by that the displacement of the posture central point of the athlete is larger than the width of the posture detection frame when the athlete lands on the ground; putting forward an anti-shielding criterion, and not positioning a landing point for a frame with shielding; the criterion condition of the shielding prevention is that if the distance from the thigh root key point to the ankle key point is smaller than a threshold value, the hip is considered to shield the foot, and the frame does not position the rear edge of the landing point; the method for accurately positioning the rear edge of the landing point is concretely implemented as follows:
after the athlete finishes landing, firstly obtaining a motion segmentation map by a KNN background difference method, obtaining the position of the athlete in a picture by the position of a human body posture detection frame, determining an effective area of a landing point by setting a long jump test area, extracting the motion segmentation map of an intersection area of the athlete posture detection frame and the long jump test area, and carrying out morphological processing on the motion segmentation map to obtain a rough segmentation map only containing feet of the athlete, namely landing parts; carrying out inverse perspective transformation on the rough segmentation image and the color original image to obtain a vertical visual angle image of the long jump test area; extracting the foot boundary of the vertical view of the rough segmentation graph, extracting the vertical view of the color graph in the corresponding boundary, carrying out bilateral filtering on the vertical view to remove the influence of the ground texture, then carrying out secondary wavelet transformation, extracting horizontal high-frequency details to obtain a detail graph of the shoe, and carrying out canny edge detection on the detail graph to obtain the edge of the shoe; calculating the intersection part of the vertical visual angle graph of the rough segmentation graph and the canny detection graph, wherein the point set which is closest to the left side, namely the jumping line is judged as the rear edge of the landing point of the athlete;
s4, measuring the actual distance from the positioning point to the jumping wire, namely the measured standing long jump score;
and S5, displaying the standing long jump state and the result measurement result of the athlete.
2. The method for measuring standing long jump performance based on video according to claim 1, wherein in S1 and S2, the captured video image is a high resolution image, specifically an image with a resolution of 3264 x 2448.
3. The video-based standing long jump performance measuring method according to claim 1, wherein in S2, human body posture information is detected in real time using a multi-person posture detector OpenPose.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111632521.5A CN114307117B (en) | 2021-12-28 | 2021-12-28 | Standing long jump result measuring method and device based on video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111632521.5A CN114307117B (en) | 2021-12-28 | 2021-12-28 | Standing long jump result measuring method and device based on video |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114307117A CN114307117A (en) | 2022-04-12 |
| CN114307117B true CN114307117B (en) | 2022-12-09 |
Family
ID=81014140
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111632521.5A Active CN114307117B (en) | 2021-12-28 | 2021-12-28 | Standing long jump result measuring method and device based on video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114307117B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114870385A (en) * | 2022-05-11 | 2022-08-09 | 安徽理工大学 | Established long jump testing method based on optimized OpenPose model |
| CN114904251A (en) * | 2022-06-01 | 2022-08-16 | 甄十信息科技（上海）有限公司 | Method and device for determining standing long jump data by wearable equipment |
| CN115761603B (en) * | 2023-01-09 | 2023-05-19 | 浙江大学计算机创新技术研究院 | Video image standing long jump distance measurement method based on improved YOLOV5 calibration-free |
| CN115845350B (en) * | 2023-03-02 | 2023-05-09 | 成都谷帝科技有限公司 | Method and system for automatic ranging of standing long jump |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110264498A (en) * | 2019-06-26 | 2019-09-20 | 北京深醒科技有限公司 | A kind of human body tracing method under video monitoring scene |
| CN110975270A (en) * | 2019-12-20 | 2020-04-10 | 西南交通大学 | Standing long jump detection method based on marks and computer vision |
| CN113137923A (en) * | 2020-01-17 | 2021-07-20 | 上海淡竹体育科技有限公司 | Standing long jump sport result measuring method |
-
2021
- 2021-12-28 CN CN202111632521.5A patent/CN114307117B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114307117A (en) | 2022-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114307117B (en) | Standing long jump result measuring method and device based on video | |
| CN111368810B (en) | Sit-up detection system and method based on human body and skeleton key point identification | |
| US9996739B2 (en) | System and method for automatic gait cycle segmentation | |
| US11182599B2 (en) | Motion state evaluation system, motion state evaluation device, motion state evaluation server, motion state evaluation method, and motion state evaluation program | |
| US11033205B2 (en) | System and method for analyzing gait and postural balance of a person | |
| US9235753B2 (en) | Extraction of skeletons from 3D maps | |
| CN107392086B (en) | Human body posture assessment device, system and storage device | |
| CN111027417B (en) | Gait recognition method and gait evaluation system based on human body key point detection algorithm | |
| CN110969114A (en) | Human body action function detection system, detection method and detector | |
| KR101118654B1 (en) | rehabilitation device using motion analysis based on motion capture and method thereof | |
| CN105913045A (en) | Method and system for counting of sit-up test | |
| CN111883229B (en) | Intelligent movement guidance method and system based on visual AI | |
| CN105999655A (en) | A counting method and system for push-up tests | |
| CN113137923A (en) | Standing long jump sport result measuring method | |
| CN112568898A (en) | Method, device and equipment for automatically evaluating injury risk and correcting motion of human body motion based on visual image | |
| CN114100103B (en) | Rope skipping counting detection system and method based on key point identification | |
| Shukla et al. | A comparison of four approaches to evaluate the sit-to-stand movement | |
| CN113780220A (en) | Child sitting posture detection method and system based on child face recognition | |
| CN116721471A (en) | Multi-person three-dimensional attitude estimation method based on multi-view angles | |
| CN116612529A (en) | Pedestrian falling behavior identification method and device based on human dynamics centroid model | |
| Serrano et al. | Automated feet detection for clinical gait assessment | |
| CN115331304A (en) | Running identification method | |
| CN115116125A (en) | Push-up examination method and implementation device thereof | |
| Zhou et al. | Kinematics parameter analysis of long jump based on human posture vision | |
| Chikano et al. | Person matching technology using gait information of 2D pose estimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |