一种新的激光雷达目标姿态估计算法

A Novel Method for Ladar Target Orientation Estimation

在激光雷达目标识别中,通常需要估计目标姿态,进而将点云和模型匹配以完成识别。对目标姿态与投影点云分布的关系进行了分析,提取了投影点云密度熵(PDE)特征对点云分布进行度量。依据目标姿态与PDE之间的关系,提出了一种新的目标姿态估计方法,将点云绕坐标轴旋转,并计算旋转后点云的PDE,以PDE最小值所对应的旋转角作为目标姿态角。仿真了5类地面装甲目标在不同视点下的激光雷达点云,对比分析了PDE方法和矩形拟合法及主成分分析(PCA)方法在自遮挡下和遮挡下的目标姿态估计性能,讨论了参数选择对算法性能的影响,给出了PDE方法的快速实现方式。实验结果表明,PDE方法在自遮挡和遮挡情况下的姿态角估计性能及算法稳健性明显优于矩形拟合法和PCA方法,特别适合严重遮挡下的目标姿态估计。

In the area of ladar target recognition,orientation estimation is always needed to align the point cloud with models for recognition.The relationship between the target′s orientation and the point cloud projection distribution is analyzed,a feature called projection density entropy（PDE） is extracted to measure this distribution.A novel method for target orientation estimation is proposed according to the relationship between the target′s orientation and PDE.The point cloud is rotated around one coordinate axis,then the PDE of the rotated point cloud is calculated,and the rotation angle with the least PDE is taken as the orientation angle.Point clouds of five kinds of ground armored targets under different viewpoints are simulated,and the performances of PDE,rectangle fitting and principle component analysis（PCA） methods under the situations of self-occlusion and occlusion are analyzed.The strategy of parameter selection is discussed,and a fast realization method of PDE is proposed.The experimental results show that the estimation performance and robustness of PDE method outperform rectangle fitting and PCA methods greatly in the cases of self-occlusion and occlusion.PDE is very suitable for application under severe occlusion.