用于三维点云表示扩展点特征直方图算法（.doc

用于三维点云表示的扩展点特征直方图算法( 庄祉昀，，孙广富 （国防科学技术大学 电子，湖南 长沙 410073） 摘要：局部特征提取在点云相关应用中具有十分重要的作用提出了一种点云的扩展点特征直方图（）描述子。本文邻域点的两两点对提出了一系列不变，特征点上构建一个局部参考坐标框架以获得特征描述子对旋转和平移的不变性接着将关键点局部邻域划分成子空间并每个子空间中的一个直方图，最后将所有直方图串联描述子。Bologna公共对描述子的性能进行了测试，多个现有算法进行了对比。结果，良好的性能，其结果现有的 关键词：；；表示；直方图 中图分类号：TN95 文献标志码：　　　文章编号： Extended Point Feature Histograms for 3D Point Cloud Representation ZHUANG Zhiyun, ZHANG Jun, SUN Guangfu (College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China) Abstract: Local feature extraction plays an important role in point cloud related applications. This paper therefore, presents an extended point feature histograms (EPFH) for 3D point cloud representation. Each point pair is first represented by several invariant pairwise point attributes. Then, a local reference frame is defined for a keypoint and the neighboring points of the keypoint is transformed into the local reference frame. These pairwise point attributes between the neighboring points and the keypoint are accumulated into several sub-features in a set of subspaces. These sub-features are finally concatenated and compressed into an overall feature descriptor. The EPFH descriptor was tested on a popular publicly available Bologna dataset and compared to several existing methods. Experimental results show that the proposed EPFH method outperforms several existing methods under different levels of noise and point cloud resolutions. Keywords: point cloud; local feature; feature representation; point sets; feature histograms. 特征提取是计算机视觉和模式识别领域的一个核心基础问题点云局部特征描述算法如何局部高鉴别力是一个解决的问题Johnson[1][2]采用点云的法向量及点的位置信息提出了一个。特征点每一个局部邻域点均可采用两个参数进行表示，其中一个参数是该邻域点到特征点法向量的距离另一个参数是该邻域点到特征点切平面的距离这两个参数值进行二维直方图统计描述子。Spin image描述子具有对刚性变换良好不变[1]。Hetzel采用包括点云深度值、表面以及曲率多个局部特征描述子[3]这些简单对视点变化具有良好的不变。Flint[4][5]对关键点与局部邻域点的法向量夹角直方图统计以THRIFT特征描述子。Chen[6]点的形状指数和法向量变化这两类信息local surface patch (LSP)特征描述子。usu[7]等通过关键点邻域点的fast point feature histograms (FPFH)特征描述子。Pang[8]体素化计算一系列3D Harr描述子，并将