复杂背景下的对象特征提取

华北电力大学硕士学位论文摘要摘要本文以犯罪现场遗留足迹为对象，阐述了图像处理和模式识别的相关知识，实现复杂背景下的对象特征提取。足迹是刑侦工作中经常用到的生物特征，针对足迹照片存在噪声干扰大，边缘不清晰，人工识别受主观因素影响等问题，提出了基于二维最大熵的类间方差、基于模糊聚类和形态学滤波的足迹特征识别方法。利用二维最大熵的类间方差算法或模糊聚类算法将图像进行阈值分割，得到鞋底各部分受力情况的粗略估计；同时结合形态学滤波和阈值面积消除法抑制噪声，并自适应地确定特征识别门限值；最后在原始图像上标识出特征区。该方法运算简单，能充分利用图像的梯度和灰度信息，有效消除噪声，提高特征区域边缘检测的准确性，为刑侦工作的足迹自动识别提供了新途径。关键词：图像处理，模式识别，最大类间方差，模糊聚类，形态学滤波ABSTRACTTakingthetreadinalibiasobject,thisthesisintroducessomeknowledgeaboutimageprocessingandpatternrecognition,andextractscharacteristicfeatureincomplexbackground.Treadfeaturerecognitionisoneofthemostimportantbiometricsincriminalinvestigationwork.Aschemewasdevelopedtocorrectlyreproducedistinct,continuousedgesanddecreasemanualintervenebasedonthemaximumvariancebetweenclusters(Otsu)methodandthealgorithmsoffuzzyC-meansclustering.Inthispaper,animprovedOtsualgorithmisproposed,whichisbasedonthetwo-dimensionalboundhistogram.FirsttheOtsumethodandthealgorithmsoffuzzyC-meansclusteringisusedtosegmenttheheavypressuresurfaceformtheimage,thenthemorphologicalfilterandareathresholdremovingmethodareappliedtofilterthesmallareaandthenoise,withanadaptivemethodtoselecttheareaextractionthreshold.Experimentalresultsshowthattheschemereproducesaccurate,smoothedgesduetotheuseofthegradientandgrayinformation,providinganewwayfortreadfeatureautomaticrecognition.FanJin(ControlScienceandControlengineering)Directedbyprof.TianPeiKEYWORDS:imageprocessing，patternrecognition，maximumvariancebetweenclusters(Otsu)，fuzzyclustering，morphologicalfilter华北电力大学硕士学位论文目录I目录中文摘要英文摘要第一章绪论.........................................................11.1课题研究的背景...................................................11.2历史回顾及国内外研究现状.........................................21.2.1足迹特征分析.................................................21.2.2图像工程概述.................................................51.2.2.1图像处理技术的发展与应用.................................71.2.2.2模式识别技术.............................................71.2.3计算机辅助足迹特征分析.......................................91.3课题的主要研究工作...........................................101.4论文章节安排.................................................10第二章图像的预处理...............................................122.1图像边缘剪切.................................................132.2彩色图像到灰度图像的转换......................................132.2.1彩色图像.....................................................132.2.2灰度图像.....................................................142.2.3彩色图像到灰度图像的转化.....................................142.3鞋底轮廓提取.................................................152.3.1滤波的技术分类...............................................152.3.2滤波器的实现.................................................172.3.2.1线性平滑滤波器..........................................172.3.2.2非线性平滑滤波器........................................182.3.2.3非线性锐化滤波器........................................182.3.3倒数梯度加权法...............................................192.4本章小结.....................................................20第三章鞋底图像的分割.............................................213.1图像分割的数学描述...........................................213.2阈值法的概念.................................................223.3基于点的全局阈值法...........................................233.3.1P-tile法......................................................23华北电力大学硕士学位论文目录II3.3.2双峰法.......................................................233.3.3时刻存储法...................................................233.3.4最小错误法...................................................243.3.5灰度直方图凹度分析法.........................................253.3.6一维灰度直方图熵法...........................................253.4基于区域的全局阈值法..........................................263.4.1灰度直方图变换法.............................................263.4.2基于灰度级的二次统计值的方法.................................273.5基于二维最大熵的类间方差分割算法..............................283.5.1二维直方图...................................................283.5.2最大类间方差法...............................................303.5.3算法的实现...................................................313.6基于模糊聚类的足迹图像分割....................................333.6.1模糊阈值分割.................................................333.6.2FCM聚类算法................................................353.6.3聚类样本的确定...............................................363.6.4聚类中心数目的确定...........................................373.7本章小结.....................................................37第四章图像的去噪处理及特征区域识别定位.........................394.1数学形态学滤波...............................................394.1.1概述.........................................................394.1.1.1膨胀与腐蚀..............................................394.1.1.2开运算和闭运算..........................................404.1.1.3击中击不中变换..........................................404.1.2数学形态学滤波处理...........................................414.2阈值面积消除.................................................424.3特征区域识别定位.............................................444.4本章小结.....................................................44第五章结论及展望.....................................................46参考文献...........................