摘要
目的激光雷达在自动驾驶中具有重要意义,但其价格昂贵,且产生的激光线束数量仍然较少,造成采集的点云密度较稀疏。为了更好地感知周围环境,本文提出一种激光雷达数据增强算法,由双目图像生成伪点云并对伪点云进行坐标修正,进而实现激光雷达点云的稠密化处理,提高3D目标检测精度。此算法不针对特定的3D目标检测网络结构,是一种通用的点云稠密化方法。方法首先利用双目RGB图像生成深度图像,根据先验的相机参数和深度信息计算出每个像素点在雷达坐标系下的粗略3维坐标,即伪点云。为了更好地分割地面,本文提出了循环RANSAC(random sample consensus)算法,引入了一个分离平面型非地面点云的暂存器,改进复杂场景下的地面分割效果。然后将原始点云进行地面分割后插入KDTree(k-dimensional tree),以伪点云中的每个点为中心在KDTree中搜索若干近邻点,基于这些近邻点进行曲面重建。根据曲面重建结果,设计一种计算几何方法导出伪点云修正后的精确坐标。最后,将修正后的伪点云与原始激光雷达点云融合得到稠密化点云。结果实验结果表明,稠密化的点云在视觉上具有较好的质量,物体具有更加完整的形状和轮廓,并且在KITTI(Karlsruhe Institute of Technology and Toyota Technological Institute)数据集上提升了3D目标检测精度。在使用该数据增强方法后,KITTI数据集下AVOD(aggregate view object detection)检测方法的AP_(3D)-Easy(average precision of 3D object detection on easy setting)提升了8.25%,AVOD-FPN(aggregate view object detection with feature pyramid network)检测方法的AP_(BEV)-Hard(average precision of bird’s eye view on hard setting)提升了7.14%。结论本文提出的激光雷达数据增强算法,实现了点云的稠密化处理,并使3D目标检测结果更加精确。
Objective Light detection and ranging(LiDAR)plays an important role in autonomous driving.Even if it is expensive,the number of equipped laser beams is still small,which results in a sparse point cloud density.The sparsity of the LiDAR point cloud makes 3D object detection difficult.The camera is another vital sensor used in autonomous driving because of the mature image recognition methods and its competitive price compared with LiDAR.However,it does not perform as well as LiDAR in 3D object detection task.To perceive the surrounding environment better,this study proposes a LiDAR data enhancement algorithm based on pseudo-LiDAR point cloud correction method to increase the density of LiDAR point cloud,thereby improving the accuracy of 3D object detection.This method has a wide application prospect because it is a general method to densify the point cloud and improve the detection accuracy,which does not depend on specific 3D object detection network structures.MethodThe algorithm can be divided into four steps.In the first step,the depth map is generated on the basis of the stereo RGB images by using depth estimation methods,such as pyramid stereo matching network(PSMNet)and Deep Pruner.The approximate 3D coordinates of each pixel corresponding to the LiDAR coordinate system are calculated according to the camera parameters and depth information.The point cloud formed by approximate 3D coordinates is usually called pseudo-LiDAR point cloud.In the second step,the ground points in the original point cloud are removed by ground segmentation method because they will disturb the surface reconstruction process in the following step.In order to improve the performance of ground segmentation,this study designs an iterative random sample consensus(RANSAC)algorithm.A register is used in the iterative RANSAC algorithm to store the points extracted in each iteration,which are planar point cloud but not ground point cloud.The register ensures that the next iteration of RANSAC will not be affected by these non-ground planar p
作者
宋绪杰
戴孙浩
林春雨
詹书涛
赵耀
Song Xujie;Dai Sunhao;Lin Chunyu;Zhan Shutao;Zhao Yao(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China;School of Traffic and Transportation,Beijing Jiaotong University,Beijing 100044,China)
出处
《中国图象图形学报》
CSCD
北大核心
2021年第5期1157-1168,共12页
Journal of Image and Graphics
基金
国家自然科学基金项目(61772066)
国家大学生创新创业训练项目(202010004010)。
关键词
激光雷达
数据增强
点云稠密化
伪点云
地面分割
3D目标检测
light detection and ranging(LiDAR)
data enhancement
point cloud densification
pseudo-LiDAR point cloud
ground segmentation
3D object detection
