CN116994330A - A volleyball sports tactical analysis method based on machine vision - Google Patents

A volleyball sports tactical analysis method based on machine vision Download PDF

Info

Publication number
CN116994330A
CN116994330A CN202310594794.8A CN202310594794A CN116994330A CN 116994330 A CN116994330 A CN 116994330A CN 202310594794 A CN202310594794 A CN 202310594794A CN 116994330 A CN116994330 A CN 116994330A
Authority
CN
China
Prior art keywords
frame
volleyball
detection
ball
track
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.)
Pending
Application number
CN202310594794.8A
Other languages
Chinese (zh)
Inventor
朱静
潘梓沛
王家创
尹邦政
尹晴
李捷平
董骥阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou University
Original Assignee
Guangzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangzhou University filed Critical Guangzhou University
Priority to CN202310594794.8A priority Critical patent/CN116994330A/en
Publication of CN116994330A publication Critical patent/CN116994330A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G06V40/23—Recognition of whole body movements, e.g. for sport training
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Human Computer Interaction (AREA)
  • Biophysics (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Psychiatry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Social Psychology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

一种基于机器视觉的排球运动战术分析方法，包括以下步骤：S1：使用Yolov5目标检测算法训练排球运动员和排球的模型，对排球视频中的排球运动员和排球进行识别，分别输出所述排球运动员和所述排球的检测框的坐标位置和检测框的外观特征矩阵；S2：通过改进的DeepSORT跟踪算法对所述排球运动员进行跟踪；S3：通过区域搜索对所述排球进行跟踪；S4：使用差帧法对所述排球运动员和所述排球的轨迹进行修补；S5：计算击球点和落地点；S6：对排球运动场景进行二维映射。本发明提供的技术方案能够快速稳定的识别排球运动中的图像信息，并呈现为平面图，便于教练进行战术策划。

A volleyball sports tactical analysis method based on machine vision, including the following steps: S1: Use the Yolov5 target detection algorithm to train volleyball players and volleyball models, identify the volleyball players and volleyballs in the volleyball video, and output the volleyball players and volleyballs respectively. The coordinate position of the volleyball detection frame and the appearance feature matrix of the detection frame; S2: Track the volleyball player through the improved DeepSORT tracking algorithm; S3: Track the volleyball through area search; S4: Use difference frames Method to repair the trajectory of the volleyball player and the volleyball; S5: Calculate the hitting point and landing point; S6: Perform two-dimensional mapping of the volleyball sports scene. The technical solution provided by the present invention can quickly and stably identify image information in volleyball and present it as a plan view to facilitate tactical planning by coaches.

Description

Volleyball sport tactics analysis method based on machine vision
Technical Field
The application belongs to the field of image processing, and particularly relates to a volleyball sport tactical analysis method based on machine vision.
Background
Volleyball sports are team ball sports which are increasingly popular with the public, wherein not only the exertion of the player's own abilities is required, but also the application of a number of tactics and strategies is required. Some tactics are prepared before some volleyball games with great significance, but how to judge whether a player on a court has the ability to contribute to the winning of the games and how to arrange the tactics needs to rely on a coach to analyze the play of teams in the past games, and further needs to continuously observe and analyze videos.
At present, some methods for detecting and tracking people exist, but the detection effect of volleyball sports videos is not good, and the problems of low accuracy, high ID exchange rate and the like exist.
Disclosure of Invention
In view of the existing problems, the present application aims to provide a volleyball sport tactical analysis method based on machine vision, which is characterized by comprising the following steps: s1: training a volleyball player and volleyball model by using a Yolov5 target detection algorithm, identifying the volleyball player and volleyball in a volleyball video, and respectively outputting the coordinate positions of the detection frames of the volleyball player and volleyball and the appearance characteristic matrix of the detection frames; s2: tracking volleyball players through an improved deep tracking algorithm; 3: tracking volleyball through area search; s4: repairing the track of volleyball players and volleyball by using a difference frame method; s5: calculating a batting point and a landing point; s6: and carrying out two-dimensional mapping on the volleyball sports scene.
Preferably, step S2 specifically includes the following steps:
s21: by passing throughDescribing the motion state of a volleyball player at a certain moment, wherein u and v respectively represent an abscissa and an ordinate of the central coordinate of a detection frame of a target, gamma represents the ratio of the width to the height of the detection frame, and h represents the height of the detection frame;The four parameters are the relative speeds of the first four parameters in the image coordinates; updating state information of a target through a Kalman filter, wherein the Kalman filter adopts a constant speed model and a linear prediction model, and the predicted value of the Kalman filter is (u, v, gamma, h);
s22: by the formula:calculating the association degree between the predicted value of the Markov distance measurement predicted by Kalman filtering and the detection value of the detector at the motion information level; wherein d (1) (i, j) represents the degree of matching of the motion state between the detected position of the jth frame and the predicted position of the ith frame, d j Representing the detected position of the j-th frame, y i Representing the predicted position of the ith frame, S i Representing a covariance matrix between the ith detected position and the average predicted position; extracting appearance characteristics of the detection frame and the prediction frame by using a characteristic extraction network at the appearance information layer, and passing through the formula:
calculating the association degree between the detection frame and the prediction frame; wherein d (2) (i, j) represents the minimum cosine distance between the jth detection frame and the ith prediction frame, r j To detect frame d j Calculating a descriptor corresponding to the appearance characteristic of the frame, provided that |r j |=1; for each successfully associated track, a repository is created which can store the last n descriptors simultaneously>Wherein n is a set super parameter;if d (2) (i, j) is less than the association threshold, then the association is considered successful; by the formula: c i,j ＝λd (1) (i,j)+(1-λ)d (2) (i, j) fusing two measurement values of the mahalanobis distance and the cosine distance in a linear weighting mode to be used as a final measurement, wherein lambda is a set weight; setting an updating time parameter and a survival period for all tracks, wherein the survival period value is +1 when the tracks are updated by using Kalman filtering once, clearing the survival period if the tracks are matched, and discarding the tracks if the survival period is greater than the updating time parameter;
s23: setting a set of unmatched tracks of volleyball playersThe center coordinates of the last detection frame of the track which is not matched after IoU matching are stored, and a unmatched track number sigma, a track number epsilon and a track number limiting parameter gamma are set; wherein, the initial values of sigma and epsilon are all 0, and the initial value of gamma is set as volleyball player target number 12. When the unmatched detection frame is judged to be a new track through 10 cycles, if epsilon < gamma, the new track is classified into a track set, and epsilon is added with 1; judging sigma if epsilon=gamma, and deleting a new track if sigma=0; if σ > 0, then the formula is passed:
wherein ,ui ,v i Is the center coordinate of the ith detection frame, s i,j For the i-th detection frame center coordinate and volleyball player unmatched track setThe minimum Euclidean distance of the center coordinates of the jth detection frame; if the matching condition is satisfied, s i,j If the track number is smaller than the set threshold value, the unmatched track is considered to be successfully matched with the new track, and the ID number of the track is given to the new track;
preferably, step S3 judges whether the detected ball is within the judgment area in the case where the volleyball can be detected in the continuous frame images: first, theDetection frame of volleyball ball detected by i frames whereinThe horizontal coordinate and the vertical coordinate of the upper left point and the horizontal coordinate and the vertical coordinate of the lower right point of the ith frame ball detection frame are respectively;Is the judgment area of the ball of the i-th frame,is the ratio of the area of the intersection area of the detection frame of the volleyball of the ith+1st frame and the judgment area to the area of the detection frame of the volleyball, and is +.>Andcalculated from the following formula:
wherein ,w b ，h b detecting the length and width of the image for the first frame, lambda being the set super parameter,/for->A detection frame for the detected sphere of frame i+1; setting track set K of ball b When->When the ball is judged to be correctly recognized, the detection frame of the ball is added>Added to ball trajectory set K b In (a) and (b); when->And if the detection is judged to be false, discarding the detection and not tracking.
Preferably, in the case where no ball is detected in the presence of the intermediate frame image, step S3 uses the region of the adjacent frame as the judgment region, then
Wherein n is the difference between the serial number of the ball detected by the frame and the serial number of the ball detected by the previous frame, the phi is a settable super parameter,detecting an area of an image for a first frame; when->When the ball is judged to be correctly recognized, the detection frame of the ball is added>Added to track set K b In (a) and (b); when->If the error detection is judged, the frame is abandoned and is not tracked.
The step S4 specifically comprises the following steps: the last detection frame before each track missing detection frame is set as follows:
Box i ＝[x i1 ,y i1 ,x i2 ,y i2 ]the first detection frame after each track missing frame is:
Box i+n ＝[x (i+n)1 ,y (i+n)1 ,x (i+n)1 ,y (i+n)2 ]where i is the frame number, x i1 ,y i1 ,x i2 ,y i2 The detection frames of the missed detection frames with the track frame serial numbers of i+m are respectively the horizontal coordinate and the vertical coordinate of the upper left point and the horizontal coordinate and the vertical coordinate of the lower right point of the detection frame:
preferably, step S5 is specifically:
setting a j volleyball player detection frame in the i frame as follows:the height of the volleyball player detection frame is +.>Selecting a ball striking point by detecting the ratio of the area where the minimum point is located to one third of the area on the volleyball player detection frame; by the formula: bk=localmin { K b Finding out the minimum value point of the ball track, wherein BK is the set of detection frames corresponding to the minimum value point, and localmin { K b The sphere trajectory set K is represented b Obtaining minimum values of center coordinates of detection frames of all balls in the ball detecting frame;
setting upDetecting the upper third area of the frame for the jth volleyball player in the ith frame, then the jth volleyball player and +.>Area of intersection->The percentage of the area of the detection frame of the ball is as follows:
wherein ,the frame number is i, which is the detection frame of the ball in BK; ioU ijmax IoU for maximum screening of ith frame ij ：IoU imax ＝max{IoU ij }, where { IoU } ij All of the ith frame }IoU ij ；
Coordinates of a center point of a lower frame of the ith frame ball detecting frameThe formula is as follows:
wherein ,for the i frame j volleyball player the abscissa of the upper left point of the detection frame,/for the j-th volleyball player>The horizontal and vertical coordinates of the lower right point of the frame are detected for the jth volleyball player in the ith frame;For the upper left point of the i-th frame sphere, the abscissa of the frame, is->The horizontal and vertical coordinates of the lower right point of the frame are detected for the ith frame ball;
when IoU ijmax When > 0, judgeIs the batting point, and the coordinates of the center point of the lower frame of the j volleyball player detection frame of the ith frame are +.>As coordinates in the corresponding two-dimensional image of the ball point; ioU ijmax When the number of the ball points is =0, the ball point is determined as the center coordinate of the lower frame of the corresponding ball detection frame +.>
Preferably, step S6 is specifically: describing coordinates g= (x, y, 1) from the course model with a planar map T To the original frame image coordinates g' = (u, v, 1) T Is mapped to: g' =hg, noted:
the beneficial technical effects of the application are as follows:
the application uses the YOLOv5 algorithm to detect the players and the balls, the algorithm has high reasoning speed and high accuracy, the accuracy of detecting the targets of the players is effectively improved, and the recognition speed is also accelerated; the improved deep SORT multi-target tracking algorithm is used for tracking the player, the algorithm is combined with motion information and appearance characteristics to track the player, a track re-matching module is added, the accuracy of tracking the player target is improved, and the probability of losing the target caused by shielding of the player is reduced; the homography transformation algorithm is used, so that the movement track of the player is more concrete, and the method is more suitable for analyzing the track of the player.
Drawings
FIG. 1 is a schematic flow chart of a volleyball sport tactical analysis method based on machine vision provided by the application;
FIG. 2 is an example of the output results of a preferred embodiment provided by the present application;
fig. 3 is a flow chart of the deep start tracking algorithm in a preferred embodiment provided by the present application.
Detailed Description
The following examples of the present application are described in detail, and are given by way of illustration of the present application, but the scope of the present application is not limited to the following examples. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.
In view of the prior art, as shown in fig. 1, the present application provides a volleyball sport tactical analysis method based on machine vision, which is characterized by comprising the following steps: s1: training a volleyball player and volleyball model by using a Yolov5 target detection algorithm, identifying the volleyball player and volleyball in a volleyball video, and respectively outputting the coordinate positions of the detection frames of the volleyball player and volleyball and the appearance characteristic matrix of the detection frames; s2: tracking volleyball players through an improved deep tracking algorithm; 3: tracking volleyball through area search; s4: repairing the track of volleyball players and volleyball by using a difference frame method; s5: calculating a batting point and a landing point; s6: and carrying out two-dimensional mapping on the volleyball sports scene.
The step S2 specifically comprises the following steps:
s21: by passing throughDescribing the motion state of a volleyball player at a certain moment, wherein u and v respectively represent an abscissa and an ordinate of the central coordinate of a detection frame of a target, gamma represents the ratio of the width to the height of the detection frame, and h represents the height of the detection frame;The four parameters are the relative speeds of the first four parameters in the image coordinates; updating state information of a target through a Kalman filter, wherein the Kalman filter adopts a constant speed model and a linear prediction model, and the predicted value of the Kalman filter is (u, v, gamma, h);
s22: by the formula:calculating the association degree between the predicted value of the Markov distance measurement predicted by Kalman filtering and the detection value of the detector at the motion information level; wherein d (1) (i, j) represents the detection bit of the j-th frameA degree of matching of the motion state between the position and the predicted position of the ith frame, d j Representing the detected position of the j-th frame, y i Representing the predicted position of the ith frame, S i Representing a covariance matrix between the ith detected position and the average predicted position; extracting appearance characteristics of the detection frame and the prediction frame by using a characteristic extraction network at the appearance information layer, and passing through the formula:
calculating the association degree between the detection frame and the prediction frame; wherein d (2) (i, j) represents the minimum cosine distance between the jth detection frame and the ith prediction frame, r j To detect frame d j Calculating a descriptor corresponding to the appearance characteristic of the frame, provided that |r j |=1; for each successfully associated track, a repository is created which can store the last n descriptors simultaneously>Wherein n is a set super parameter; if d (2) (i, j) is less than the association threshold, then the association is considered successful; by the formula: c i,j ＝λd (1) (i,j)+(1-λ)d (2) (i, j) fusing two measurement values of the mahalanobis distance and the cosine distance in a linear weighting mode to be used as a final measurement, wherein lambda is a set weight; setting an updating time parameter and a survival period for all tracks, wherein the survival period value is +1 when the tracks are updated by using Kalman filtering once, clearing the survival period if the tracks are matched, and discarding the tracks if the survival period is greater than the updating time parameter;
s23: setting a set of unmatched tracks of volleyball playersThe center coordinates of the last detection frame of the track which is not matched after IoU matching are stored, and a unmatched track number sigma, a track number epsilon and a track number limiting parameter gamma are set; wherein, the initial values of sigma and epsilon are all 0, and the initial value of gamma is set as volleyball player target number 12. Detection of a mismatchWhen the frame is judged to be a new track through 10 cycles, if epsilon < gamma, the new track is classified into a track set, and epsilon is added with 1; judging sigma if epsilon=gamma, and deleting a new track if sigma=0; if σ > 0, then the formula is passed:
wherein ,ui ,v i Is the center coordinate of the ith detection frame, s i,j For the i-th detection frame center coordinate and volleyball player unmatched track setThe minimum Euclidean distance of the center coordinates of the jth detection frame; if the matching condition is satisfied, s i,j If the track number is smaller than the set threshold value, the unmatched track is considered to be successfully matched with the new track, and the ID number of the track is given to the new track;
step S3, judging whether the detected volleyball is in a judging area or not under the condition that volleyball can be detected in continuous frame images: volleyball detection frame that ith frame detected whereinThe horizontal coordinate and the vertical coordinate of the upper left point and the horizontal coordinate and the vertical coordinate of the lower right point of the ith frame ball detection frame are respectively;Judging area for the ball of the i-th frame, < >>Is the ratio of the area of the intersection area of the detection frame of the volleyball of the ith+1st frame and the judgment area to the area of the detection frame of the volleyball, and is +.> andCalculated from the following formula：
wherein ,w b ，h b detecting the length and width of the image for the first frame, lambda being the set super parameter,/for->A detection frame for the detected sphere of frame i+1; setting track set K of ball b When->When the ball is judged to be correctly recognized, the detection frame of the ball is added>Added to ball trajectory set K b In (a) and (b); when->And if the detection is judged to be false, discarding the detection and not tracking.
Step S3, in the case that no ball is detected in the intermediate frame image, using the area adjacent to the frame as the judgment area, then
Wherein n is the difference between the serial number of the ball detected by the frame and the serial number of the ball detected by the previous frame, the phi is a settable super parameter,detecting an area of an image for a first frame; when->When the ball is judged to be correctly recognized, the ball is thenDetection frame->Added to track set K b In (a) and (b); when->If the error detection is judged, the frame is abandoned and is not tracked.
The step S4 specifically comprises the following steps: the last detection frame before each track missing detection frame is set as follows:
Box i ＝[x i1 ,y i1 ,x i2 ,y i2 ]the first detection frame after each track missing frame is:
Box i+n ＝[x (i+n)1 ,y (i+n)1 ,x (i+n)1 ,y (i+n)2 ]where i is the frame number, x i1 ,y i1 ,x i2 ,y i2 The detection frames of the missed detection frames with the track frame serial numbers of i+m are respectively the horizontal coordinate and the vertical coordinate of the upper left point and the horizontal coordinate and the vertical coordinate of the lower right point of the detection frame:
the step S5 specifically comprises the following steps:
setting a j volleyball player detection frame in the i frame as follows:the height of the volleyball player detection frame is +.>Selecting a ball striking point by detecting the ratio of the area where the minimum point is located to one third of the area on the volleyball player detection frame; by the formula: bk=localmin { K b Finding out the minimum value point of the ball track, wherein BK is the set of detection frames corresponding to the minimum value point, and localmin { K b The sphere trajectory set K is represented b Obtaining minimum values of center coordinates of detection frames of all balls in the ball detecting frame;
setting upDetecting the upper third area of the frame for the jth volleyball player in the ith frame, then the jth volleyball player and +.>Area of intersection->The percentage of the area of the detection frame of the ball is as follows:
wherein ,the frame number is i, which is the detection frame of the ball in BK; ioU ijmax IoU for maximum screening of ith frame ij ：IoU imax ＝max{IoU ij }, where { IoU } ij IoU for the ith frame ij ；
Coordinates of a center point of a lower frame of the ith frame ball detecting frameThe formula is as follows:
wherein ,for the i frame j volleyball player the abscissa of the upper left point of the detection frame,/for the j-th volleyball player>The horizontal and vertical coordinates of the lower right point of the frame are detected for the jth volleyball player in the ith frame;For the upper left point of the i-th frame sphere, the abscissa of the frame, is->The horizontal and vertical coordinates of the lower right point of the frame are detected for the ith frame ball;
when IoU ijmax When > 0, judgeIs the batting point, and the coordinates of the center point of the lower frame of the j volleyball player detection frame of the ith frame are +.>As coordinates in the corresponding two-dimensional image of the ball point; ioU ijmax When the number of the ball points is =0, the ball point is determined as the center coordinate of the lower frame of the corresponding ball detection frame +.>
The step S6 specifically comprises the following steps: describing coordinates g= (x, y, 1) from the course model with a planar map T To the original frame image coordinates g' = (u, v, 1) T Is mapped to: g' =hg, noted as:
the foregoing describes in detail preferred embodiments of the present application. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the application without requiring creative effort by one of ordinary skill in the art. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by a person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.

Claims (7)

1. A volleyball sport tactics analysis method based on machine vision, which is characterized by comprising the following steps:
s1: training a volleyball player and volleyball model by using a Yolov5 target detection algorithm, identifying volleyball players and volleyball in volleyball video, and respectively outputting coordinate positions of detection frames of the volleyball players and the volleyball and appearance feature matrixes of the detection frames;
s2: tracking the volleyball player by an improved deep start tracking algorithm;
s3: tracking the volleyball through area searching;
s4: repairing the volleyball player and the volleyball track by using a difference frame method;
s5: calculating a batting point and a landing point;
s6: and carrying out two-dimensional mapping on the volleyball sports scene.
2. The machine vision-based volleyball sport tactical analysis method of claim 1, wherein the step S2 specifically comprises the steps of:
s21: by passing throughDescribing the motion state of the volleyball player at a certain moment, wherein u and v respectively represent the abscissa and the ordinate of the central coordinate of a detection frame of a target, gamma represents the ratio of the width to the height of the detection frame, and h represents the height of the detection frame;The four parameters are the relative speeds of the first four parameters in the image coordinates; updating state information of a target through a Kalman filter, wherein the Kalman filter adopts a constant speed model and a linear prediction model, and the predicted value of the Kalman filter is (u, v, gamma, h);
s22: by the formula: d, d (1) (i,j)＝(d j -y i ) T S i -1 (d j -y i ) Calculating the association degree between the predicted value of the Markov distance measurement predicted by Kalman filtering and the detection value of the detector at the motion information level; wherein d (1) (i, j) represents the degree of matching of the motion state between the detected position of the jth frame and the predicted position of the ith frame, d j Representing the detected position of the j-th frame, y i Representing the predicted position of the ith frame, S i Representing covariance between the i-th detected position and the average predicted positionA matrix; extracting appearance characteristics of the detection frame and the prediction frame by using a characteristic extraction network at the appearance information layer, and passing through the formula:
calculating the association degree between the detection frame and the prediction frame; wherein d (2) (i, j) represents the minimum cosine distance between the jth detection frame and the ith prediction frame, r j To detect frame d j Calculating a descriptor corresponding to the appearance characteristic of the frame, provided that |r j |=1; for each successfully associated track, a repository is created which can store the last n descriptors simultaneously>Wherein n is a set super parameter; if d (2) (i, j) is less than the association threshold, then the association is considered successful; by the formula: c i,j ＝λd (1) (i,j)+(1-λ)d (2) (i, j) fusing two measurement values of the mahalanobis distance and the cosine distance in a linear weighting mode to be used as a final measurement, wherein lambda is a set weight; setting updating time parameters and survival time limit for all tracks, wherein the survival time limit value is +1 when the tracks are updated by using Kalman filtering once, clearing the survival time limit if the tracks are matched, and discarding the tracks if the survival time limit is greater than the updating time parameters;
s23: setting a set of unmatched tracks of the volleyball playerThe center coordinates of the last detection frame of the track which is not matched after IoU matching are stored, and a unmatched track number sigma, a track number epsilon and a track number limiting parameter gamma are set; wherein, the initial values of sigma and epsilon are all 0, and the initial value of gamma is set as the goal number 12 of the volleyball player. When the unmatched detection frame is judged to be a new track through 10 cycles, if epsilon < gamma, the new track is classified into a track set, and epsilon is added with 1; judging sigma if epsilon=gamma, and deleting a new track if sigma=0; if sigma is greater than 0,then the formula is passed:
wherein ,ui ,v i Is the center coordinate of the ith detection frame, s i,j For the set of the center coordinates of the ith detection frame and the unmatched track of the volleyball playerThe minimum Euclidean distance of the center coordinates of the jth detection frame; if the matching condition is satisfied, s i,j And if the track number is smaller than the set threshold value, the unmatched track is considered to be successfully matched with the new track, and the ID number of the track is assigned to the new track.
3. The machine vision based volleyball sport tactical analysis method of claim 2, wherein the step S3, in the case where the volleyball can be detected in the continuous frame image, judges whether the detected volleyball is within the judgment area: volleyball detection frame that ith frame detected whereinThe horizontal coordinate and the vertical coordinate of the upper left point and the horizontal coordinate and the vertical coordinate of the lower right point of the ith frame ball detection frame are respectively;Judging area for the ball of the i-th frame, < >>Is the ratio of the area of the intersection area of the detection frame of the volleyball of the ith+1st frame and the judgment area to the area of the detection frame of the volleyball, and is +.> andCalculated from the following formula:
wherein ,w b 、h b respectively detecting the length and width of the image of the first frame, wherein lambda is a set super parameter,A detection frame for the detected sphere of frame i+1; setting track set K of ball b When->When the ball is judged to be correctly recognized, the detection frame of the ball is added>Added to ball trajectory set K b In (a) and (b); when->And if the detection is judged to be false, discarding the detection and not tracking.
4. The machine vision based volleyball sport tactics analysis method of claim 3, wherein in the case where no ball is detected in the presence of the intermediate frame image, the step S3 uses the area adjacent to the frame as the judgment area, then
Where n is the frame detectedThe difference between the serial number of the ball and the serial number of the ball detected in the previous frame, the phi can be set as a super parameter,detecting an area of an image for a first frame; when->When the ball is judged to be correctly recognized, the detection frame of the ball is added>Added to track set K b In (a) and (b); when->If the error detection is judged, the frame is abandoned and is not tracked.
5. The machine vision based volleyball sport tactical analysis method of claim 4, wherein the step S4 is specifically: the last detection frame before each track missing detection frame is set as follows:
Box i ＝[x i1 ,y i1 ,x i2 ,y i2 ]the first detection frame after each track missing frame is:
Box i+n ＝[x (i+n)1 ,y (i+n)1 ,x (i+n)1 ,y (i+n)2 ]where i is the frame number, x i1 ,y i1 ,x i2 ,y i2 The detection frames of the missed detection frames with the track frame serial numbers of i+m are respectively the horizontal coordinate and the vertical coordinate of the upper left point and the horizontal coordinate and the vertical coordinate of the lower right point of the detection frame:
6. the machine vision based volleyball sport tactical analysis method of claim 5, wherein the step S5 is specifically:
setting the j-th volleyball player detection frame in the i-th frame as follows:the height of the volleyball player detection frame is +.>Selecting a ball striking point by detecting the ratio of the area where the minimum point is located to the third area on the volleyball player detection frame; by the formula: bk=local min { K b Finding out the minimum value point of the ball track, wherein BK is the set of detection frames corresponding to the minimum value point, and local min { K b The sphere trajectory set K is represented b Obtaining minimum values of center coordinates of detection frames of all balls in the ball detecting frame;
setting upFor the third area on the detection frame of the j-th volleyball player in the i-th frame, the j-th volleyball player and the +.>Area of intersection->The percentage of the area of the detection frame of the ball is as follows:
wherein ,the frame number is i, which is the detection frame of the ball in BK; ioU ijmax IoU for maximum screening of ith frame ij ：IoU imax ＝max{IoU ij }, where { IoU } ij IoU for the ith frame ij ；
The center point of the lower frame of the ith frame ball detecting frameCoordinates ofThe formula is as follows:
wherein ,for the i frame, the j th said volleyball player detects the abscissa of the upper left point of the frame,The horizontal and vertical coordinates of the lower right point of the volleyball player detection frame are the ith frame and the jth frame;For the upper left point of the i-th frame sphere, the abscissa of the frame, is->The horizontal and vertical coordinates of the lower right point of the frame are detected for the ith frame ball;
when IoU ijmax When > 0, judgeIs the batting point, and the coordinate of the center point of the lower frame of the detection frame of the volleyball player on the ith frame and the jth frame is +.>As coordinates in the corresponding two-dimensional image of the ball point; ioU ijmax When the number of the ball points is =0, the ball point is determined as the center coordinate of the lower frame of the corresponding ball detection frame +.>
7. The machine vision based volleyball sport of claim 6The tactical analysis method is characterized in that the step S6 specifically comprises: describing coordinates g= (x, y, 1) from the course model with a planar map T To the original frame image coordinates g' = (u, v, 1) T Is mapped to: g' =hg, noted as:
CN202310594794.8A 2023-05-24 2023-05-24 A volleyball sports tactical analysis method based on machine vision Pending CN116994330A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310594794.8A CN116994330A (en) 2023-05-24 2023-05-24 A volleyball sports tactical analysis method based on machine vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310594794.8A CN116994330A (en) 2023-05-24 2023-05-24 A volleyball sports tactical analysis method based on machine vision

Publications (1)

Publication Number Publication Date
CN116994330A true CN116994330A (en) 2023-11-03

Family

ID=88525527

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310594794.8A Pending CN116994330A (en) 2023-05-24 2023-05-24 A volleyball sports tactical analysis method based on machine vision

Country Status (1)

Country Link
CN (1) CN116994330A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI913186B (en) * 2025-06-10 2026-01-21 國立虎尾科技大學 Volleyball match real-time tactical analysis system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114549577A (en) * 2022-02-17 2022-05-27 浙江工业大学 Volleyball movement track detection and restoration method based on deep learning
CN114820702A (en) * 2022-04-14 2022-07-29 大连理工大学 A Pedestrian Multi-target Tracking Method Based on yolov5 in Deepsort UAV Perspective

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114549577A (en) * 2022-02-17 2022-05-27 浙江工业大学 Volleyball movement track detection and restoration method based on deep learning
CN114820702A (en) * 2022-04-14 2022-07-29 大连理工大学 A Pedestrian Multi-target Tracking Method Based on yolov5 in Deepsort UAV Perspective

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
PAN ET.AL: "Research on volleyball players tracking based on improved DeepSORT", 《2022 4TH INTERNATIONAL CONFERENCE ON COMMUNICATIONS, INFORMATION SYSTEM AND COMPUTER ENGINEERING (CISCE)》, 17 August 2022 (2022-08-17), pages 591 - 595 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI913186B (en) * 2025-06-10 2026-01-21 國立虎尾科技大學 Volleyball match real-time tactical analysis system

Similar Documents

Publication Publication Date Title
Datta et al. Person-on-person violence detection in video data
CN111444890A (en) Sports data analysis system and method based on machine learning
US11707663B1 (en) System for tracking, locating and predicting the position of a ball in a game of baseball or similar
US20250029387A1 (en) A System for Tracking, Locating and Calculating the Position of a First Moving Object in Relation to a Second Object
CN111905350A (en) Automatic table tennis hitting performance evaluation method and system based on motion data
Ren RETRACTED ARTICLE: A novel approach for automatic detection and identification of inappropriate postures and movements of table tennis players: W. Ren
CN116797887A (en) Football training assistance system and method based on football videos
CN114973409B (en) Method and system for identifying goal scoring based on court environment and personnel pose
Ashfaq et al. Badminton player’s shot prediction using deep learning
Terroba et al. Finding optimal strategies in tennis from video sequences
He et al. Notice of violation of ieee publication principles: study on sports volleyball tracking technology based on image processing and 3D space matching
Moshayedi et al. Kinect based virtual referee for table tennis game: TTV (Table Tennis Var System)
Athanesious et al. Perspective transform based YOLO with weighted intersect fusion for forecasting the possession sequence of the live football game
Li Tactical analysis of table tennis video skills based on image fuzzy edge recognition algorithm
CN116994330A (en) A volleyball sports tactical analysis method based on machine vision
Li et al. Analytical model of action fusion in sports tennis teaching by convolutional neural networks
Jannet et al. A Deep Learning Approach to Badminton Player Footwork Detection Based on YOLO Models: A Comparative Study
He et al. Mathematical modeling and simulation of table tennis trajectory based on digital video image processing
Chakraborty et al. Deep learning-based prediction of football players’ performance during penalty shootout
CN116486297B (en) A method and system for predicting shooting accuracy by combining pose determination and force estimation
Wang Low-cost Badminton Trajectory Recognition and Landing Point Prediction Optimization Based on Field Coordinate System Transformation
CN118053208A (en) Badminton video action recognition method and device, electronic equipment and storage medium
Nelikanti et al. An optimization based deep lstm predictive analysis for decision making in cricket
Meng Deep Learning Algorithm and Video Image Processing-based Basketball Training System
Song et al. Research and analysis of table tennis movement trajectory prediction model based on deep learning

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20231103