CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleGround Object Tracking Based on Aerial Images
Thesis Advisor杨一平 ; 田原
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机应用技术
Keyword目标跟踪 Mean-shift 斜框统计量 主对称轴 凹凸区域 离散尺度 一定尺度下的二阶方向导数图 Target Tracking Mean-shift Statistics Of Oblique Rectangle Principal Axis Of Symmetry Concave And Convex Region Discrete Scale Two Directional Derivative Under Certain Scale
Abstract智能视觉在无人机侦察、导航、精确打击中有重要应用。本文针对航拍图像序列下地面运动目标的跟踪问题展开研究,主要的创新点如下: 1、针对航拍图像序列中车辆的姿态变化较大,使用传统的水平垂直的固定矩形跟踪框会把较多的背景像素统计到目标物上,造成目标表达的不精确的问题,本文研究了据车辆姿态使用倾斜矩形框内的颜色统计量表达目标物,并对车辆主对称轴进行跟踪,从而提高了目标模型的描述精度,提高了跟踪的稳定性和跟踪精度。 2、针对在使用整个目标物的特征统计量时,目标物的表达不够精细,跟踪鲁棒性不够的问题,本文提出了使用目标物上的多个相对均匀的块状区域内的特征统计量表达目标物。具体做法是,根据图像曲面的属性,对图像曲面进行块状区域划分。并利用目标物上多块区域内的颜色统计量及面积对目标物进行跟踪,实现了使用多个局部区域块内的特征对目标物的跟踪,有一定的鲁棒性。 3、为了确定目标物的尺度,在基于图像曲面属性分割的基础上,提出了凸显跟踪目标的离散尺度的计算方法。进而生成在相应尺度下的以二阶方向导数为权重的置信图,并结合mean-shift方法跟踪目标。
Other AbstractGround object tracking is a very important part in intelligent unmanned aerial vehicles, which is vital to vision-based surveillance, vision-based navigation and accurate strikes. This thesis explores the tracking problem in aerial images. The main innovating points are as follows: (1) We have a better statistics of the initial target. Compared to the traditional rectangle statistics, we compute the statistics of the target using oblique rectangle according to the posture of the target in aerial images. What’s more, we track the principal axis of the target, so that we improve the precision and robustness in tracking. (2) The tracked object forms curved surface through projection. This thesis divides the curved surface of the target into block areas by the property of its points. We track the object using the RGB statistics and the area of the block areas, so that we implement tracking using local regions. This method has certain robustness. (3) It does not consider the scale of the object in Chapter 3. Here we provide a simple method to choose scale of the tracked object. Then we implement tracking by mean shift method using confidence map which is formed of two directional derivative under the computed scale.
Other Identifier200828014629074
Document Type学位论文
Recommended Citation
GB/T 7714
李海昌. 基于航拍图像序列的地面目标跟踪[D]. 中国科学院自动化研究所. 中国科学院研究生院,2011.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20082801462907(1416KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李海昌]'s Articles
Baidu academic
Similar articles in Baidu academic
[李海昌]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李海昌]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.