中文摘要：

提出了一种基于特征匹配的三维自动目标识别方法．首先使用均匀采样选择特征点和自旋图描述子提取特征，定义相似度度量并且使用最近邻方法得到初始的匹配；其次使用向量场一致性算法消除错误的匹配；最后根据算法保留的正确匹配数目进行目标识别．针对目标识别的实际应用需求，进一步研究了点云的空间分辨率、激光雷达测距误差对目标识别性能的影响，可以为激光雷达三维目标识别系统的设计提供参考．

英文摘要：

On the basis of feature correspondence, an automatic target recognition algorithm was pres- ented from 3D imaging LIDAR. Firstly, uniform sampling strategy and spin-image descriptor were a- dopted. The similarity metric was defined and nearest neighbor algorithm was used to determine the initial correspondence. Then the vector field consensus algorithm was used to remove mismatches, and the resulted matches were used for target recognition. For practical applications, the influence of target spatial resolution and the noise level on target recognition performance were investigated. The result of the study provides the reference for parameter design of the 3D target recognition system.