【作者】 付向南；

【导师】 张树功；

【作者基本信息】 吉林大学 ， 计算数学， 2008， 硕士

【摘要】 本文综述了静止背景下,介绍运动目标跟踪的应用背景,理论与方法,文章主要是对人的运动图像序列的分析处理,对所涉及到的运动目标检测,运动目标分类,目标跟踪几部分进行了详细的阐述,并列举一些常见的算法,分类介绍各种算法的优缺点,并讨论一些具体问题的优化技术,如遮挡问题,阴影问题,实时性问题等。最后总结视觉跟踪的未来发展前景及努力的研究方向。更多还原

【Abstract】 Computer Vision is an independent discipline,Computer Vision’s goal is to make a computer through images ( compared static images, sports images include more information) have a ability of Cognitiving the surrounding environment information,This capability not only makes the computer can sense objects’geometric information, including its shape,location,posture, movement,but also describes them,storage,identiflcation and understanding. Movement of visual analysis in the field of computer vision in recent years been concerned,Based on the Vision Campaign Analysis,one of the potential applications is visual monitoring,which not only has a high scientific value but also has a huge potential economic value.This paper is about motion analysis .There are three questions: motion detection,classification and the target tracking.Moving target detection’s purpose is extracting change image sequences from the background image.Several of these methods including: 1) background subtraction 2) Difference between frames, the frame difference Improvement Act 3) optical flow,background image estimation methods including:Statistical Average,IIR Filter, Adaptive Gaussian Background Model;there are many threshold selection method.Having the threshold segmentation algorithm,Ostu algorithm,dual-threshold method,and so on.Target detection steps can be divided into 1) preconditioning 2) Background Model 3) prospects for detection 4) post-processing.Object pretreatment methods: Neighborhood average method,median filtering method,the paper also describes the mathematical morphology.Target Classification will get an unknown target from observation datas, with a group of suitable characteristics as input and outputing target category or the possibility of different types of objectives.This paper briefly describes the objective classification:based on the shape of information classification, based on the characteristics of sports classification.Moving target tracking in the image sequence needs to get a target location in each image,based on the goals and objectives of the environment in which,select one or more targets characteristics which can distinguish target from the background.Moving target tracking can not only provide the trajec-tory,as well as provide a reliable source of data for moving target Scene Analysis and Motion Analysis,and moving target tracking information has helped in the correct moving target detection and identification of target movement. For tracking process, Scene assumptions, Moving Target Type, the campaign features moving target, cameras and camera movement, deformation objects, background changes, irregular objects Campaign in the campaign process and the color and shape, the tracking methods change vary widely. The difficulty of tracking is how fast and reliable images match goal from one frame to another image.Fast can guarantee real-time tracking, and reliability of tracking is the most basic requirement. Moving target tracking method can be summarized as follows: 1) Based on the regional matching,which combined with least squares fitting and template matching,will be effective in reducing the amount of computation,improving real-time.2) Based on the feature matching.3) Based on the model matching.Finally,various methods in this article are compared and the paper discusses a number of specific issues,such as shielding,shadows,real-time optimiza- tion technology and so on,and summarizes the visual tracking development directions and prospects.更多还原