摘要
普通激光雷达的扫描方式通常都是固定角分辨率,这样就导致了远距离物体表面上的扫描点变得稀疏,进而在点云配准时由于重合度不高造成配准精确度的下降,而且远距离的小物体容易被漏扫描;而与之相反,当满足远距离扫描要求时,近距离物体表面上的扫描点将异常稠密,造成了大量的数据冗余,需要进行点云压缩。为了解决采样不均带来的问题,一种基于核密度估计并根据期望空间分辨率自动调整激光雷达扫描角分辨率的算法被提出。在激光雷达原型系统上的实验表明了其能自动检测待扫描物体,并根据物体的距离调整扫描步进以达到期望的空间分辨率,并在扫描测量精度上超过了普通激光雷达系统。
The scanning pattern of light detection and ranging(LiDAR)systems is usually taken with a fixed angular step,consequently the sample points of the object at long distances are sparse,and thus sparsity of point clouds always leads to small overlap,it will provide an inaccurate point clouds registration,meanwhile,small objects which at long distance easily be missing.On the other hand,when it meets the requirements of long distances,the sample points of closed objects will be too dense,lead to abundantly redundant data,which need to be compressed.To address the issues of unbalance sampling,a scanning strategy is proposed to adaptively adjust the space resolution depending on the kernel density estimation of the objects being scanned.Experiments on prototype system verify its ability to detect objects,then to adjust its scanning step according to object's distance to achieve expected space resolution,its measurement accuracy surpass common LiDAR system.
作者
李树青
林靖宇
LI Shuqing;LIN Jingyu(School of Electrical Engineering,Guangxi University,Nanning 530004)
出处
《计算机与数字工程》
2023年第5期988-992,共5页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61561005)
国家留学基金项目(编号:201906660002)
广西研究生教育创新计划项目(编号:YCSW2019026)资助。
关键词
激光雷达
采样不均
空间分辨率
核密度估计
LiDAR
unbalance sampling
space resolution
kernel density estimation
