基于尺度不变特征变换方法的物体检测

基于尺度不变特征变换方法的物体检测1基于尺度不变特征变换方法的物体检测摘要图像匹配时指通过一定的算法在两幅或者多幅图像之间识别特征点的过程。图像匹配技术是图像处理方面极为重要的研究内容，是图像信息领域之中的一项重要的技术。与此同时，它也是图像分析技术的基础，其应用性极为广泛。图像匹配是一个由多个步骤组成的过程，不同的匹配算法之间的步骤也会有很大的不同，效果也有较大的区别。SIFT算法是一种基于空间尺度的算法，这个算法的特征是图像的局部特征，使得其具备旋转不变性，尺度不变性，对于视角的变化和噪声也有比较好的稳定性。本课题研究了基于SIFT算法的图像匹配，阐述了SIFT算法的详细实现过程，通过实验，检测了SIFT算法在尺度变化，旋转变化下的图像匹配效果。并运用SIFT算法在自建数据库下实现了物体识别定位。先在简单背景下进行物体识别定位，再在复杂背景下进行尝试。最后对于立体物体进行尝试检测。关键字：图像匹配，尺度不变性，旋转不变性，SIFT算法，物体识别基于尺度不变特征变换方法的物体检测2OBJECTDETECTIONBASEDONSCALEINVARIANTFEATURETRANSFORMMETHODABSTRACTReferstotheprocesswhenacertainimagematchingalgorithmbetweentwoormoreimagesbyidentifyingthefeaturepoints.Imagematchingtechnologyisextremelyimportantimageprocessingresearchisanimportantfieldofinformationtechnologyamongtheimages.Atthesametime,itisalsothebasisofimageanalysistechniques,theapplicationofveryextensive.Imagematchingisaprocessconsistingofseveralsteps,thestepbetweenthedifferentmatchingalgorithmswillhaveverydifferenteffectshaveagreaterdifference.SIFTalgorithmisbasedonthespatialscaleofthealgorithm,whichischaracterizedbylocalfeaturesoftheimage,makingitwithrotationalinvariance,scaleinvariance,theperspectivechangesandnoisealsohavebetterstability.ThispaperstudiedtheimagematchingalgorithmbasedonSIFT,elaboratedindetailSIFTalgorithmimplementationprocess,throughexperiments,testingtheeffectofimagematchingalgorithmSIFTinscalechange,rotationchange.Andtheuseofself-builtdatabaseundertheSIFTalgorithmtoachievetheobjectrecognitionpositioning.Keywords：ImageMatching，Scaleinvariance，Rotationalinvariance，SIFTalgorithm，ObjectRecognition基于尺度不变特征变换方法的物体检测3目录1.绪论.............................................................................................................................11.1课题背景..............................................................................................................11.2课题现状..............................................................................................................11.3主要内容.............................................................................................................31.4章节安排.............................................................................................................32SIFT算法原理...........................................................................................................52.1SIFT算法介绍.....................................................................................................52.2DOG尺度空间构造.............................................................................................52.3特征点搜索和定位.............................................................................................82.4方向赋值.............................................................................................................92.5特征点描述........................................................................................................102.6SIFT特征向量匹配...........................................................................................112.7本章小结...........................................................................................................113.部分SIFT代码分析................................................................................................123.1DOG尺度空间的构建....................................................................................123.2关键点的搜索与定位.......................................................................................123.3方向赋值...........................................................................................................143.4关键点描述.......................................................................................................153.5特征点的匹配...................................................................................................163.6本章小结...........................................................................................................174.利用SIFT算法在自建数据库下图像检测............................................................184.1检测环境............................................................................................................184.2图像特征点提取................................................................................................184.3特征点匹配检测结果........................................................................................20基于尺度不变特征变换方法的物体检测44.3.1尺度变换下的匹配.....................................................................................204.3.2旋转变换下的匹配.....................................................................................224.4物体识别...........................................................................................................234.4.1物体识别与定位思路.................................................................................244.4.2自建图像数据库.........................................................................................244.4.3简单背景下测试结果.................................................................................264.4.4复杂背景下的测试结果.............................................................................284.5近似平面物体与立体物体检测.....................