Learning Parameters of a Recognition System Based on Local Affine Frames.pdfVIP

下载本文档

2
0
约 8页
2017-04-13 发布于江苏
举报
版权申诉

Learning Parameters of a Recognition System Based on Local Affine Frames.pdf

1、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
4、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
5、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
6、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
7、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Learning Parameters of a Recognition System Based on Local Affine Frames

Cognitive Vision Workshop (CogVis02) September 2002, Zurich, Switzerland Learning Parameters of a Recognition System Based on Local Affine Frames Jir??? Matas 1,2 and S?te?pa?n Obdrz?a?lek 1,2 1 Center for Machine Perception, Czech Technical University, Prague, 120 35, CZ 2 Centre for Vision Speech and Signal Processing, University of Surrey, Guildford, GU2 7XH, UK Abstract An approach to object recognition, based on matching of local image features, is presented. First, distinguished regions of data-dependent shape are robustly detected. On these regions, local affine frames are established using sev- eral affine invariant constructions. Direct comparison of photometrically normalised colour intensities in local, geo- metrically aligned frames results in a matching scheme that is invariant to piecewise-affine image deformations, but still remains very discriminative. Nevertheless, invariance to a wide range of local ge- ometric and photometric transformations reduces the dis- criminative power – not all possible transformations are equiprobable. Probability of the transformations is esti- mated from matches established by the invariant method on the training data. The estimate is exploited in the recogni- tion phase to favour local correspondences with more likely transformations. The potential of the approach is experimentally verified on COIL-100 – a publicly available image database. 99.9% recognition rate is obtained for 18 training views per object. 1 Introduction Our interest is in an object recognition system which is able to recognise 2D and 3D objects using a representation learned from a given training set of labelled images of the objects. A recognition system, plausibly aspiring to at least partially emulate the capabilities of humans, shall have a number of desirable attributes. It should ? be able to learn the object representation from only a few examples. ? be stable under viewpoint and illumination changes. ? be robust to occlusion and background c