采用投影轮廓特征的激光雷达快速目标识别

激光雷达可以获得目标的三维形状信息,已成为目标识别领域新的研究热点。针对传统匹配识别算法计算量大的问题,提出了一种快速的激光雷达目标识别方法。采用由粗到精的策略,提出了一种新的点云正交投影轮廓特征(PCF)实现模型的快速预选,在此基础上利用迭代最近点(ICP)算法将目标与模型点云精确匹配,并综合利用特征匹配和点云匹配信息构建相似性度量实现目标识别。采用25类地面装甲目标在96个不同视角下的点云数据进行实验,结果表明该算法的运算效率远优于逐一匹配法,且对目标姿态估计误差和目标遮挡具有很强的稳健性,具有较好的综合性能和应用推广价值。

Fast Target Recognition in Ladar Using Projection Contour Features

Ladar has became a hot research topic in target recognition community for its ability to capture the object′s explicit 3D shape. To avoid the heavy computation burden faced by the traditional recognition methods, a fast ladar target recognition method is proposed. Adopting the coarse-to-fine strategy, the models are quickly pre-screened by using the proposed orthogonal projection contour features (PCF). Then the target is registered with the hypothesis models precisely by using the iterative closest point (ICP) method. Finally, the similarities between the target and models are calculated by integrating the features matching and points matching results. Therefore, the target is recognized to the model who owes the highest similarity. The experiments are performed using the point-clouds of 25 ground targets under 96 different viewing angles. The results show that the proposed method outperforms the existing methods greatly in the term of computation efficiency, and it is very robust to orientation error and occlusion. Thus, this method is the state-of-the-art and have a high application potential.