目标跟踪是目前图像处理、模式识别和计算机视觉等领域的重要研究内容，它的目的是在一组图像序列中持续可靠地确定目标的位置。本文研究了基于。NET平台EmguCV图像处理框架下，利用Mean-Shift（均值漂移）算法实现动态跟踪视频序列中移动目标的问题。软件完成了EmguCV图形库下的视频播放及图像处理功能，在此基础上提取视频图像特征建立目标模型的直方图。本文还涉及对跟踪目标的自适应的Mean-Shift算法分析，针对当运动目标增大或减小时跟踪框无法实时地适应运动目标外观变化的情况，将优化后Mean-Shift算法添加到跟踪系统中。经过几个视频实验，结果显示在EmguCV框架下Mean-Shift算法执行时跟踪精度高、稳定性好，可以应用于实时系统。87424

毕业论文关键词   目标跟踪；Mean-Shift算法；尺度自适应；EmguCV

毕业设计说明书外文摘要

Title    Adaptive tracking window of video tracking algorithm 

             research and EmguCV framework implementations       

Abstract

Target tracking is an important research content in the fields of image processing, pattern recognition and computer vision。 The goal is to determine the location of the target consistently and reliably in a set of image sequence。 This paper mainly studies how to realize dynamic tracking of moving objects in video sequences, based on the 。NET platform EmguCV image processing framework and using Mean-Shift algorithm。 This software accomplished the video playback and image processing functions under the EmguCV graphics library。 On the basis of this, the feature of video image is extracted and the histogram of the target model is established。 This paper also involves the analysis of the adaptive Mean-Shift algorithm for tracking target, for example, when the moving target increases or decreases, the tracking frame can not adapt to the change of the appearance of the moving object in real time。 So the optimized Mean-Shift algorithm is added to the tracking system。 After several video experiments, the results show that the Mean-Shift algorithm has high tracking accuracy and good stability in the EmguCV framework, it can be used in real time system。

Keywords   Target tracking；Mean-Shift；Scale adaptation；EmguCV

目   次

1  绪论 1

1。1  算法研究意义 1源-于Y优+尔-论.文:网www.chuibin.com 原文+QQ7520^18766

1。2  图像处理框架 2

1。3  章节安排 2

2  Mean-Shift算法 3

2。1  研究现状 3

2。2  Mean-Shift算法基本思想及其扩展 3

2。3  Mean-Shift算法目标跟踪架构 7

3  EmguCV图形库的视频处理平台 9

3。1  EmguCV介绍 9

3。2  EmguCV图像处理 9

3。3  EmguCV视频框架 11

4  自适应Mean-Shift跟踪算法 15

4。1  仿射模型和自适应Mean-Shift算法 15

4。2  软件框架 17

4。3  算法功能模块及流程 19

5  实验结果 24

结论