航空影像辅助下的城区机载LiDAR汽车目标检测方法

发布时间：2018-04-26 12:43

本文选题：激光雷达滤波 + 车辆检测　；参考：《遥感技术与应用》2014年05期

【摘要】：机载激光雷达(LiDAR)技术的出现为地面汽车目标检测提供了新的途径。为了从机载LiDAR点云数据中提取汽车对象,根据不同地物的属性特征,提出了一种航空影像辅助下的城区机载LiDAR汽车目标检测方法。首先利用形态学开重建滤波完成地面和地物的分类,然后在地物点的基础上结合正射影像,通过归一化植被指数(NDVI)特征完成对植被和非植被地物的初步分类,最后在非植被地物的基础上,根据地物对象的形状特征及高程信息完成汽车和建筑物及阴影植被等非汽车对象的分类,从而完成汽车目标的提取工作。3个实验区的计算结果表明:该方法能有效从LiDAR点云中提取汽车目标,正确度和完整度的均值分别为95%和85%,满足实用性要求。
[Abstract]:The emergence of airborne lidar provides a new approach for ground vehicle target detection. In order to extract vehicle objects from airborne LiDAR point cloud data, a method of airborne LiDAR vehicle target detection in urban area assisted by aerial image was proposed according to the attribute features of different ground objects. Firstly, the classification of ground and ground objects is completed by morphological reconstruction filtering, and then the primary classification of vegetation and non-vegetation features is completed by using the normalized vegetation index (NDVI) feature based on the orthophoto image. Finally, on the basis of non-vegetation features, the classification of non-vehicle objects such as vehicles, buildings and shadow vegetation is completed based on the shape characteristics and elevation information of the objects. The results of three experimental areas show that the method can extract vehicle targets effectively from the cloud of LiDAR points. The mean values of accuracy and completeness are 95% and 85% respectively, which meet the practical requirements.
【作者单位】：西南交通大学地球科学与环境工程学院;
【基金】：国家自然科学基金项目(51178404、41072220) 高等学校博士学科点专项科研基金(20100184110019) 中央高校基本科研业务费专项资金(SWJTU09CX010、SWJTU09BR050)资助
【分类号】：TN958.98

【参考文献】