一种基于激光三维成像雷达距离像的目标检测方法  

A target detection method based on lidar 3D range image

作　　者：黄明晶[1] 蹇渊 王雪梅[1] 孟伟杰 马蒙蒙 HUANG Ming-jing;JIAN Yuan;WANG Xue-mei;MENG Wei-jie;MA Meng-meng(Southwest Institute of Technical Physics,Chengdu 610047,China)

机构地区：[1]西南技术物理研究所,四川成都610047

出　　处：《激光与红外》2020年第7期803-807,共5页Laser & Infrared

摘　　要：为了提高激光三维成像雷达地面目标的检测效率,提出一种基于激光三维成像雷达距离像的目标检测方法。该方法首先采用基于邻域像素内检测距离反常算法对距离像进行噪声抑制预处理,然后采用基于形态学的地面估计和高程分割算法实现了距离像的地物分割,最后根据感兴趣目标的尺寸特征,采用最小外接矩形估计算法实现了感兴趣目标的快速检测。本文提出的方法充分利用了距离像的高程信息和目标已知的先验知识,不受二维图像中光照灰度变化对目标检测效果的影响以及处理三维海量点云数据对计算速率的影响。实验结果证明,本文提出的方法适用于复杂多变的战场环境下对地面目标进行快速检测,且满足了实时性要求。In order to improve the efficiency of target detection in LIDAR 3D range Image,it is proposed a target detection method based on lidar 3D range image.This method firstly restrain the noise for range image within abnormal distance detection algorithm in the neighborhood pixels,then the morphological ground estimation and height segmentation algorithm is used to complete the ground object segmentation of range image,finally according to the size characteristics of the interesting target,it gets the interesting target fleetly by Minimum Area Bounding Rectangle estimation algorithm.It makes full use of height information and prior information known to the target,the efficiency of target detection is unacted on variation of light intensity for 2D image,computing speed is unacted on processing 3D massive point cloud data.The results show the method is suitable for rapid ground target detection in the complex and varied battlefield environment.

关 键 词：激光三维雷达 距离像 形态学算法 最小外接矩形 目标检测 

分 类 号：TN958.98[电子电信—信号与信息处理]