Meanshift视频监控系统中地面目标稳定跟踪算法研究
毕业论文关键词 目标跟踪 均值漂移 光流 粒子滤波
Title Resasrch of Ground-target Stable TrackingAlgorithm in Video Surveillance System
Abstract As a important means of video information storage and traceability, videosurveillance technology has played an increasingly important role in the fieldof military and civilian control.However, due to the huge amount of videoinformation, it is tedious to get information quickly from a lot if vedio.Especially when in the certain occasions such as an unattended occasion, we oftenneed the video surveillance system to automatically extract, identify and trackthe moving target. Therefore, it is necessary to carry out the research of automatictarget recognition technology based on video stream. This topic mainly carry outresearch work for technology and methods of moving target detection under thecamera motion state. There are three main difficulties: 1)Target appearance modelchanges greatly in a short time resulted by perspective transformation; 2)Wrongtracking resulted by sample region which is similar to target feture in a complexscenario;3)Target is in the half or full occlusion state, which leads to loss oftracking. To solve the above problems, this topic focuses on moving targetdetection and tracking based on the principle of meanshift as well as that basedon the principle of optical flow and particle filter. In this topic, I use theMatlab software to simulate these three algorithms to achieve the effect of movingtarget detection and tracking, then analyse and compare their performance suchas accuracy，robustness and speed. This topic proposed that the main point of futureresearch is to solve the problem of wrong tracking in the Meanshift algorithm.
Keywords Target Tracking Meanshift Optical Flow Particle Filter
目 次
1 绪论 1
1.1 课题背景及研究意义 1
1.2 课题研究现状与发展趋势 1
1.3 运动目标检测跟踪技术研究中存在的问题 2
1.4 本文主要内容与结构安排 3
2 基于 Meanshift 的运动目标检测跟踪算法 5
2.1 基于 Meanshift 的运动目标检测跟踪算法原理及实现 5
2.2 基于Meanshift的运动目标检测跟踪算法的改进 10
3 基于光流的运动目标检测跟踪算法原理及实现 12
4 基于粒子滤波的运动目标检测跟踪算法原理及实现 15
4.1 目标模型描述 15
4.2 粒子样本集合描述 16
4.3 基于颜色直方图的粒子滤波算法 16
5 算法性能分析与比较 19
5.1 实验结果 19
5.2 性能分析 21
结论 22