| With the popularity of mobile devices and the development of mobile software,people are increasingly interested in processing and analyzing videos on the mobile side.Moreover,due to the continuous improvement of the performance of the hardware assembled on the mobile device,such as the frame rate and resolution supported during video shooting,the main frequency of the processor,the number of cores,etc.,the processing efficiency and quality of video on the mobile side are also increasingly satisfying users.As people pay more and more attention to sports health,mobile sports teaching analysis software is also emerging in an endless stream.Using mobile side software to track a specific target and draw the target’s trajectory is very important for users to analyze whether the motion in the video is standard.This thesis will focus on target tracking technology,combined with the advantages of deep learning technology in the field of computer vision,design and implement a single target tracking model with high performance,and integrate it into the mobile video analysis system to track and plot the trajectory of the target.The thesis first analyzes and summarizes the development status of single target tracking and key point detection technology in recent years.Based on this,discusses the requirements analysis of target tracking algorithm and mobile video analysis system,and focuses on the functional and non-functional requirements involved;then the thesis deeply analyzes the single target tracking algorithm and motion trajectory fitting algorithm,and evaluates them.Then,according to the demand analysis,the thesis describes the general design and detailed design of the mobile video analysis system,introduces the module division of the system,namely four modules of video recording,video management,video analysis and network communication,and explains how to integrate the key algorithms into the video analysis module.Finally,introduces the functional and nonfunctional tests of the system,and analyzes the test results. |
