激光雷达地面目标姿态估计

激光雷达距离像能表征目标的几何结构,而同一目标不同姿态的距离图像差别较大,因此目标姿态的准确估计对目标识别尤为重要。Ulf Grenander的模式理论利用变形模板表示观测图像,目标姿态变化引起刚性模板的变形用特殊正交群矩阵表示,在贝叶斯估计框架下,用跳扩散过程从后验密度中取样,以希尔伯特-施密特范数作为误差特性,对目标姿态进行估计,并分析误差性能。针对激光雷达距离像基于地面目标进行研究,利用Grenander的模式理论,在不同载噪比情况下对车辆姿态进行估计。通过对姿态估计误差的分析,表明该算法能够较准确地估计目标的姿态。

Laser radar for ground-based target orientation estimation

Range imaging of laser radar can reflect the targets geometry structure, and different poses of the same target lead to a tremendous variety of range imaging. Thus pose estimation is especially important for automated target recognition (ATR). Ulf Grenander's pattern theory makes use of deformable template to represent observed image and model the variability of target pose via the actions of the matrix special orthogonal group. In Bayesian estimation framework, it samples from posterior to empirically generate the estimates using jump-diffusion process. Error analysis of pose estimation makes use of the Hilbert-Schmidt norm as an error metric. Our work focuses on pose estimation of ground-based targets of range imaging, using Ulf Grenander's pattern theory to estimate vehicle in differe