结合兴趣点和边缘的建筑物和物体识别方法-北京大学智能科学系.pdf

第 18 卷 第 8 期 计算机辅助设计与图形学学报 Vol 18 , No8 2006 年 8 月 JOURNAL OF COMPU TERAIDED DESIGN COMPU TER GRAPHICS Aug , 2006 结合兴趣点和边缘的建筑物和物体识别方法 王君秋 　　　查红彬 (北京大学视觉与听觉信息处理国家重点实验室　北京　100871) (jer ywang @public3 bta net cn) 摘 要 　提出了多种图像特征相结合的建筑物和物体识别方法 使用尺度不变特征描述器描述的 HarrisLaplace 兴 趣点以及边缘颜色直方图描述的边缘特征表示图像 边缘和兴趣点包含图像的重要信息 对 2 种特征的抽取同时进 行 :基于 Harris 检测器可以直接得到边缘特征 ;在多个尺度下进行 Harris 兴趣点检测 ,利用 Laplace 公式得到 Harris Laplace 兴趣点 进行物体识别时 ,根据兴趣点的数目自适应地改变兴趣点和边缘特征的相似性权重 与同类方法相 比较表明,该方法具有更高的识别正确率 ,在视点变化、光照条件变化等情况下具有较好的性能 关键词 　建筑物识别 ;物体识别 ;兴趣点 ;边缘 ; 图像检索 中图法分类号　TP39141 Combining Interest Points and Edges for Building and Object Recognition Wang J unqiu 　Zha Hongbin ( ) N ational L aboratory on Machi ne Perception , Peki ng U niversity , Beiji n g 　100871 Abstract 　　We present a novel approach using combined features to recognize images containing specific objects and buildin gs The content of an image is characterized by two kinds of features : HarrisLaplace in terest points described by the scale invariant feature transform ( SIFT) descriptor and edges described by the edge color histogram Edges and corners contain the maximal amount of information necessary for object reco gnition The feature detection in this work is an integrated process : edges are detected directly based on the Harris function ; Harris interest points are detected at several scales and HarrisLaplace interest points are found using the Laplace function The combination of edges and interest points brings efficient feature detec tion and high recognition ratio to the object reco gnition system Experimental results show that this system has good performance Key words 　building recognition ; object reco gnition ; interest points ; ed