摘要
在介绍二维目标检测方法的基础上,对基于深度学习的三维目标检测方法中的深度神经网络展开探讨,包括间接处理、直接处理和融合处理3类基本方法,并着重分析和对比各深度神经网络在三维目标检测速度和精确度等方面的优缺点,为车载激光雷达目标检测方法的选择提供参考依据。
Based on the introduction of 2D-object detection methods,the deep neural network for 3D-object detection based on deep learning,including 3 principal methods of indirect processing,direct processing,and fusion processing,are discussed.The advantages and disadvantages of deep neural network in 3D-object detection speed and accuracy are analyzed and compared,which provide reference basis for the selection of object detection method of onboard laser radar.
作者
彭育辉
郑玮鸿
张剑锋
Peng Yuhui;Zheng Weihong;Zhang Jianfeng(Fuzhou University,Fuzhou 350116)
出处
《汽车技术》
CSCD
北大核心
2020年第9期1-7,共7页
Automobile Technology
基金
福建省科技厅产学合作重大项目(2017H6007)。
关键词
无人驾驶
深度学习
激光雷达
目标检测
三维点云
Autonomous driving
Deep learning
Laser radar
Object detection
3D point cloud