摘 要

运动目标检测与跟踪是计算机视觉领域的热点问题，广泛应用于视频监控、人机交互、虚拟现实和图像压缩等。而要在各种复杂的环境中和不同的条件下（如遮挡、光照变化等）都对目标进行准确的跟踪是目前广大研究工作者共同关注的焦点，也是目前实际应用中一个亟待解决的难题。

本文主要研究静态背景下运动目标的检测、运动目标跟踪以及相关结果仿真分析三方面的内容。运动目标检测方面，在分析了目前比较常用的三种目标检测方法，即光流法、帧间差分法、背景相减法的基础上，着重研究了基于帧间差分法运动目标检测的算法原理及流程，讨论了三种检测算法的优缺点。运动目标跟踪方面，在分析了目前比较常用的三种目标跟踪算法，即均值漂移算法、卡尔曼滤波算法、基于特征的目标跟踪算法的基础上，重点研究了基于特征——最小外接矩形框运动目标跟踪算法。分析了其算法原理以及跟踪步骤。最后用matlab软件采用帧间差分运动目标检测法以及基于最小外接矩形框跟踪法对含有运动目标的视频进行仿真。

在采用了帧间差分检测法以及最小外接矩形框跟踪法基础上，用matlab软件对视频进行仿真，检测到了人体的轮廓，同时矩形框跟踪出了运动人体的轨迹，达到了运动目标检测与跟踪的效果。

关键词：运动目标检测；运动目标跟踪；帧间差分法；最小外接矩形框

Abstract

Moving target detection and tracking is a hot issue in the field of computer vision, is widely used in video surveillance, human-computer interaction, virtual reality and image compression etc.. But in various complex environments and different conditions (such as occlusion, illumination changes) of target accurate tracking is currently the focus of researchers of common concern in the current application, is an urgent problem to be solved.

This paper mainly involves moving target detection under a static background, moving target tracking and simulation results analysis of three aspects. In the moving object detection, in the analysis of the current commonly used three kinds of target detection method based on optical flow method, namely, the inter-frame difference method, background subtraction method, focusing on the frame difference algorithm principle and process method based on moving object detection. And discusses the advantages and disadvantages of three kinds of detection algorithm. In terms of moving object tracking, in the analysis of the current three kinds of target more commonly used tracking algorithm, the mean shift algorithm, Calman algorithm, based on the characteristics of the target tracking algorithm based on the characteristics, focuses on the minimum exterior rectangle based on moving target tracking algorithm. Finally using MATLAB software using frame difference detection method for moving targets as well as based on the minimum bounding box tracking method to simulate the video with moving objects.

Based on the frame difference detection method and the minimum bounding box tracking method based on the simulation, the video with the MATLAB software, to detect the contours of the body, at the same time rectangle tracking a human motion trajectory, reached the moving target detection and tracking results.

Keywords: moving object detection; object tracking; inter-frame difference method; minimum circumscribed rectangle

目 录

1  绪论 1

1.1  研究背景和意义 1

1.2  国内外研究现状 2

1.3  章节安排 4

2  运动目标检测与跟踪技术 5

2.1  数字图像处理相关概念 5

2.1.1  数字图像处理过程 5

2.1.2  图像增强 5

2.1.3  图像分割 6

2.1.4  数学形态学 6

2.2  运动目标检测流程及常用算法 7

2.2.1  背景差分法 8

2.2.2  帧间差分法 8

2.2.3  光流法 9

2.3  运动目标跟踪常用算法

2.3.1  基于均值漂移目标跟踪算法

2.3.2  基于卡尔曼滤波目标跟踪算法

2.3.3  基于特征的目标跟踪算法

3  基于帧间差分法运动目标检测的研究

3.1  帧间差分法运动目标检测流程

3.2  帧间差分法运动目标检测过程及原理

3.2.1  RGB图像转换为灰度图像

3.2.2  图像差分处理

3.2.3  差分图像二值化

3.2.4  形态学滤波 16

3.2.5  连通性检测 17

4  基于最小外接矩形框目标跟踪 19

4.1  目标跟踪流程 19

4.1.1  运动目标跟踪过程图 19

4.1.2  运动目标跟踪过程分析 19

4.2  基于最小外接矩形框跟踪原理

4.2.1  特征提取

4.2.2  最小外接矩形提取

4.3  最小矩形框跟踪实现

5  仿真结果与分析

5.1  仿真环境

5.2  运动目标检测仿真

5.3  运动目标跟踪仿真 28

结 论 33

致 谢

参考文献

附录A 38

附录B 54

附录C 67