理工大学学士学位论文PAGE\*MERGEFORMATIII学士学位论文基于红外图像的目标特征提取方法研究摘要红外图像目标特征提取是红外预警系统与成像制导的核心技术之一。红外系统的隐蔽性好，抗干扰能力强，作用距离远，搜索速率快，无镜面回波，可昼夜工作，这些优良特性使其越来越多地应用于军事和民用领域中。因此，如何提取鲜明的特征来提高红外图像目标的检测、识别性能以及整个红外系统的智能化水平一直是国内外学者的研究方向和焦点。本文旨在研究图像特征提取方法，使得能在分类识别中得到较高的识别率。首先，简要介绍了红外图像分析、传统提取方法分析，主要是Hu不变矩的理论，研究它在提取图像特征所表现出的特性。其次，针对传统提取方法的弊端，研究利用小波不变矩的特征提取方法。实验结果表明，采用小波不变矩特征提取方法可获得较高的识别结果，从而也满足了实际系统的实时性要求。虽然在针对旋转图像时需要进行归一化，但这并不影响它在未来科技中的发展。关键词：红外图像；特征提取；Hu不变矩；小波不变矩AbstractFeatureextractionofinfraredimagestargetsareoneofthekeytechnologiesininfraredearlywarningandimagingguidance.Infraredsystemshavegoodconcealment,stronganti-interferenceability,faroperatingrangeandfastsearchspeed.Besides,theyhavenomirrorechoandcanworkdayandnight.Alloftheseadvantagesmakethemusedinmilitaryorcivilianfieldsmoreandmore.Consequently,howtoextractthedistinctfeatures,toimprovethedetectionandrecognitionofinfraredimagesandthustoimprovetheintelligentleveloftheoverallinfraredsystems,hasbeentheresearchdirectionandfocalpointforthedomesticandforeignscholars.Thispaperfocusesonthemethodsofimagefeatureextractionandgettinghigherrecognitionratefortheclassificationtasks.Firstly,brieflyintroducedtheinfraredimageanalysis,analysisofthetraditionalextractionmethods,mainlyHuinvariantmomenttheory,studiesthepropertiesexhibitedbytheimagefeatureextraction.Secondly,aimingatthedisadvantagesoftraditionalextractionmethods,featureextractionmethodbasedonwaveletinvariantmoments.Theexperimentalresultsshowthat,usingthewaveletmomentinvariantfeatureextractionmethodcangethigherrecognitionresults,whichalsomeetsthereal-timerequirementsofactualsystem.Althoughthenormalizationintherotatedimageneed,butthisdoesnotaffectitinthefuturedevelopmentofscienceandtechnology.Keywords:infraredpicture;featureextraction;Huinvariantmoments;waveletinvariantmomentsPAGE\*MERGEFORMATIII目录TOC\o"1-3"\h\z\uHYPERLINK\l_Toc73321绪论PAGEREF_Toc73321HYPERLINK\l_Toc248101.1研究背景及意义PAGEREF_Toc248101HYPERLINK\l_Toc201