CASIA OpenIR  > 毕业生  > 硕士学位论文
Thesis Advisor马颂德
Degree Grantor中国科学院自动化研究所
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Abstract目标物体识别.即从图象中提取目标物体实例,是计算机视觉中的一个重要问 题。从80年代末开始,基于索引的识别方法受到广泛重视。此法的基本思想就是 用一种表查找机制取代计算复杂度非常高的传统的特征匹配机制。所有基于索引的 目标物体识别系统都是计算图象的不变量作为索引关键字,从haLsh表中寻找模型 编号。本文研究以图象的几何特征为不变量的几何h2Lshing法。研究主要集中在用 加权机制改进原始的几何hashing以及几何hashing与另一种基于表查找机制的识别 技术:Hough变换的类比两方面。 几何hashing法的致命弱点在于景物噪声的存在导致不变量的无法精确求取。误 选、漏选是主要应该避免的问题。Rigoutsos和Hummel用概率模型重新解释了几 何hashing法.并在此基础上实现了相似变换下的最优加权几何hashing。本文第二 和第三章将把上述方法扩展到仿射变换的情况。 作为另一种基于表查找机制的目标物体识别方法,广义Hough变换十几年来得 到充分的发展,特别难得的是广义Hough变换抗干扰的能力很强,这恰好是HaLshing 法的最弱点。为了更好地把已经相当成熟的广义Hough变换技术应用到hashing法 中,本文第五章将对二者做详细比较.并推导出它们本质上的一致性。 在实验过程中,作者归纳了若干条启发式规则,这些规则对一个真正实用的目 标物体识别系统来讲是必要的。
Other AbstractObject recognition is a central problem in computer vision. It involves a set of object models that must be recognized in images. Index-based object recognition methods have received great attention since the end of 1980's, whose main idea is to replace the traditional high-computational-complexity matching method with computational effiecient indexing method. All index-based object recognition systems use invariants as keywords to look for models in the hash-table. This thesis devotes to geometric hashing which takes advantage of image's geometric properties. The research focuses on developing the original geometric hashing with weighted voting scheme and the comparison between geometric hashing and another important table-lookup-based object recognition scheme: Hough transformation. The fatal drawback of geometric hashing is that noise in the scenary leads to the impossiblity of getting highly reliable invariants. False alarm and false negative are main issues which should be avoided. A new explanation of geometric hashing based on Bayesian probability model was proposed by Rigoutsos and Hummel, and a weighted- voting scheme was presented. The second chapter and the third chapter of this thesis will develop this scheme to the environment which could recognize objects undergone affine transformation. Genetic Hough Transformation has received great development in the lastest ten to twenty years. The most distinguished attribute of this method is its high resistablility to noise, which is just the shortcoming of geometric hashing. In order to apply the well- developed Hough transformation technology to hashing, the fifth chapter gives the detailed comparison between them and claims that they are unifiable in essence. Some heuristic rules are also deduced in this thesis, which are important in an applicable recognition system.
Other Identifier432
Document Type学位论文
Recommended Citation
GB/T 7714
张磊. 目标物体识别中的几何Hashing[D]. 中国科学院自动化研究所. 中国科学院自动化研究所,1996.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张磊]'s Articles
Baidu academic
Similar articles in Baidu academic
[张磊]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张磊]'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.