基于激光三维成像数据的车辆目标方位估计  

Vehicle Pose Estimation Based on 3D Laser Imaging Data

作　　者：闫小伟[1] 邓甲昊[1] 孙志慧[1] 

机构地区：[1]北京理工大学机电工程与控制国防科技重点实验室．北京100081

出　　处：《科技导报》2009年第8期39-42,共4页Science & Technology Review

基　　金：国防科技重点实验室项目(9140C3601130802);兵器预研支撑基金项目(62301110404);北京理工大学研究生科技创新项目(1020011140301)

摘　　要：对目标方位进行精确估计是提高目标匹配识别效率的重要前提,而车辆目标的地面投影轮廓近似于矩形,为此提出了基于激光三维成像数据的车辆目标方位估计方法。首先依据激光成像数据的高程值对数据进行分类,从激光成像区域的三维数据中提取出车辆目标数据进而获取目标的投影轮廓数据,然后采用向量积异号排除法求解车辆目标投影轮廓数据的凸边界,在此基础上采用直接求解凸边界最小外接矩形的方法对车辆目标的方位进行初步估计,最后利用与目标投影轮廓数据邻近的地面数据,并依据最小二乘原理拟合优化凸边界的最小外接矩形,校正车辆目标初始估计方位,从而实现车辆目标方位的精确估计。仿真结果表明,该方法可实现车辆目标方位的快速准确估计。The target pose estimation can be used to improve the efficiency of the target matching recognition. Most vehicles＇ edges are approximately of rectangular shape in a top view. In view of that fact, a method for vehicle pose estimation based on 3D laser imaging data is proposed. The first step is to extract the vehicle object data from the laser imaging data by using data height classification method, to obtain vehicles＇ edges data. Then an exclusive method based on the sign of the vector product is used to identify the convex edges among the vehicles＇ edges data. Based on the convex edges, a rectangle with minimal area is obtained . Through such a rectangle, the original pose of the vehicle can be estimated. The minimal rectangle estimate and the estimated pose of vehicle are eventually improved by using least squares method in terms of both vehicle object data and ground data in vicinity of the vehicle object. So, a better estimation can be made. The resuhs show that the vehicle pose estimation can thus be obtained with high precision and in real-time.

关 键 词：目标识别 激光成像 方位估计 最小二乘法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]