基于自旋图的三维自动目标识别

Three dimensional automatic target recognition based on spin-images

将三维表面匹配领域的自旋图(spin-images)方法运用于扫描激光雷达自动目标识别中,基于自旋图特征,提出了平均最大相似性度量、候选对应关系数目、几何一致性筛选后对应关系数目、分组筛选后最大组的对应关系数目这4个识别指标,并将其组合构成组合识别准则,从而进行三维目标识别。同时针对实际应用需求,进一步研究了点云空间分辨率、激光雷达测距误差以及遮挡对目标识别率的影响,为激光雷达三维目标识别系统的设计提供了参考。

Scanning LIDAR automatic target recognition based on spin-images method was proposed which was used in three dimensional surface match area.The four criterions were proposed,including the average maximum similarity measure,the number of candidate correspondences,the number of correspondences after geometric consistency filtering,the correspondences number of maximum group after group filtering correspondence.The combined four criterions were constituted a combined recognition criterion for target recognition.In view of the actual application requirements,the influences of point cloud spatial resolution,LIDAR ranging noise and occlusion on target recognition rate were further studied.The study in this paper can provide reference for parameter design of the three dimensional target recognition system.