深度信息辅助的视频目标跟踪算法研究和实现

摘要目标跟踪技术一直是计算机视觉领域的研究热点。由于物体的漂移、遮挡和周遭环境变化等原因，同时也一直是研究的难点。传统方法在RGB视频中进行目标的跟踪，但是，RGB数据容易受到光照条件变化等各种因素影响。深度数据相对于彩色数据来说，包含了空间信息，有利于处理目标跟踪问题中的一些难点问题。所以，本文将使用Kinect传感器所获得的深度和RGB图像来研究视频目标跟踪，还将利用深度信息处理在目标跟踪过程中出现的遮挡问题。此外，本文对该算法进行了进一步的优化，使得跟踪效果更加鲁棒。最后，本文在普林斯顿RGB-D数据集上验证算法，并与一些传统算法进行比较。通过观察所得出的实验结果，证明了本文的算法具有更高的准确率和鲁棒性。49136

关键词深度信息目标跟踪遮挡处理

毕业论文设计说明书外文摘要

Title Study and implement of visual object tracking algorithms based on depth information

Abstract

In the field of computer vision,object tracking technology has been the hot spot of research.Because of the model drift,occlusion and the environment changes,the technology has always been the difficulties of the research.The traditional methods for object tracking do research in RGB videos.However,RGB data is easily affected by various factors such as the change of light conditions.Relative to the color data,the depth data contains spatial information,which is beneficial to deal with some difficult problem in the object tracking.So in this paper,We will use depth information and RGB images obtained from Kinect to study object tracking algorithm in videos,we will also use the depth information to handle the occlusion problems in the process of tracking.Besides,this paper will optimize the algorithm which makes the tracking effect robust.Finally,this paper will run this algorithm on the Princeton RGB-D Tracking dataset,compared with some traditional algorithm.By the observations,the experimental results show that the algorithm of this paper has higher accuracy and robustness.

Keywords RGBD object tracking occlusion handling

1 引言 1

1.1 研究背景 1

1.2 研究现状与难点 1

1.3 研究内容 3

1.4 论文组织结构 3

2 目标跟踪基础算法 4

2.1 获取RGB-D图像 4

2.2 彩色和深度图像信息处理 5

2.3 低维压缩特征 6

2.4 算法步骤 8

2.5 本章小结 9

3 遮挡问题处理机制 10

3.1 建立模型 10

3.2 遮挡判断算法 12

3.3 目标恢复算法 14

3.4 修正目标位置 15

3.5 本章小结 16

4 目标跟踪算法优化 17

4.1 图像噪声处理 17

4.2 多尺度分析 17

4.3 本章小结 20

5 目标跟踪系统深度信息辅助的视频目标跟踪算法研究和实现:http://www.youerw.com/jisuanji/lunwen_52049.html