基于视频图像处理的罚篮技术动作辅助训练系统

发布时间：2018-04-12 14:45

本文选题：罚篮 + 特征提取　；参考：《浙江理工大学》2017年硕士论文

【摘要】：罚篮是运动员重要的训练项目,也是比赛中最为有效的得分方式。由于其技巧性强,所以需要从最初的动作开始训练,然后循序渐进的养成一种适合自己的投篮节奏。目前辅助运动员罚篮训练的方式还停留在使用摄像机记录投篮视频,然后通过观看录像回放的方式来指导运动员,该方式依靠的是教练的经验和运动员的主观来判断每次投篮的得失,缺乏科学的指导,也容易忽视运动员的个体特殊性。针对上述问题,论文中应用视频图像的相关算法,设计了一套辅助运动员罚篮训练的系统。系统会自动分析运动员的动作、技术参数、出手信息,并给出相应的辅助意见。下面是论文的主要研究内容:(1)在运动员的罚篮视频中,存在许多的罚篮无关帧,真正的罚篮帧集包含了起始、抬臂、下蹲、伸展的所有过程。为了获取罚篮帧集和球的出手信息,论文中通过分析多个样本的多次罚篮视频提出了出手帧检测算法,建立了出手帧数学模型,可以快速精准的定位出手帧,确定罚篮帧集,为进一步的获取篮球的出手状态信息做准备。并通过多个样本的实验,验证了算法的可靠性和鲁棒性。(2)在罚篮过程中对篮球的分析至关重要。针对罚篮视频中篮球的特征,论文在标准的Hough变化的基础上,提出了篮球定位检测算法。一方面,针对视频处理的特性,提取篮球前景和边缘信息对干扰像素点进行剔除,减少计算次数。另一方面,从圆参数方程入手,将像素点空间映射到圆参数空间,再降低累加器维度,使用多维数组代替多重循环进行投票,获取圆半径,计算出圆心坐标。通过实验对比验证了算法的有效性和抗干扰性。(3)在出手帧检测和篮球定位算法的基础上,获取出手速度与角度。为了矫正罚篮时运动员的持球手型,提出了手型获取算法,获取手型,提供给运动员参考。通过对专业运动员的罚篮技术进行分析总结,概括了罚篮动作的技术要领,并对其进行参数量化,最终指导运动员的动作。(4)系统平台设计。系统的前台用户交互界面使用Qt编写,包含系统的初始化和相应的功能菜单。后台的图像处理模块使用c/c++编程,调用OpenCV2.4.9的视频图像库来实现。数据处理模块主要实现对手轨迹信息的筛选、出手信息的计算、技术参数的获取,并把相应的数据存放到MySQL数据库中。
[Abstract]:Free throw is an important training item for athletes, and it is also the most effective way to score in competitions.Because of its skill, it needs to start from the initial action, and then gradually develop a rhythm suitable for their own shooting.At present, the way to assist athletes in free throw training is to use cameras to record shooting videos, and then to guide athletes by watching video playback.This way depends on the experience of coach and the subjective of athletes to judge the gain and loss of each shot, lacks scientific guidance, and easily ignores the individual particularity of athletes.In order to solve the above problems, this paper designs a system to assist athletes' free throw training by using the video image algorithm.The system will automatically analyze the athlete's movement, technical parameters, release information, and give the corresponding auxiliary advice.The following is the main research content of the thesis: (1) in the athletes' free throw video, there are many free basket independent frames. The real set of free throw frames contains all the processes of starting, lifting arms, squat, and stretching.In order to obtain the shot information of the free throw frame set and the ball, the paper puts forward the shot frame detection algorithm by analyzing the multiple samples of the free throw frame video, and establishes the mathematical model of the shot frame, which can locate the shot frame quickly and accurately, and determine the free throw frame set.Prepare for further information on the shooting status of basketball.The reliability and robustness of the algorithm are verified by the experiments of several samples, and it is very important to analyze basketball in the process of free throw.Aiming at the basketball features in the free throw video, this paper puts forward a basketball location detection algorithm based on the standard Hough change.On the one hand, according to the characteristics of video processing, basketball foreground and edge information are extracted to eliminate the interference pixels and reduce the number of calculations.On the other hand, starting with the circular parameter equation, the pixel space is mapped to the circular parameter space, and then the accumulator dimension is reduced, and the multi-dimensional array is used to vote instead of multiple loops to obtain the radius of the circle and calculate the coordinates of the center of the circle.Experimental results show that the algorithm is effective and anti-jamming. (3) on the basis of shot frame detection and basketball positioning algorithm, the speed and angle of shot are obtained.In order to correct the player's hand type in free throw, an algorithm of hand type acquisition is put forward, which can be used as a reference for athletes.By analyzing and summarizing the technique of free throw for professional athletes, this paper summarizes the technical essentials of the action of free throw, quantifies its parameters, and finally guides the design of system platform for athletes' action.The front user interface of the system is written by QT, including the initialization of the system and the corresponding function menu.The background image processing module is programmed by c / c, and the video image library of OpenCV2.4.9 is called.The data processing module mainly realizes the selection of the opponent trajectory information, the calculation of the release information, the acquisition of the technical parameters, and stores the corresponding data in the MySQL database.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：G841;TP391.41

【参考文献】