基于深度学习的无人驾驶汽车目标检测算法研究  

Research on Target Detection Algorithms for Autonomous Vehicles Based on Deep Learning

作　　者：朱思瑶 申彩英[1] ZHU Siyao;SHEN Caiying(Automobile and Traffic Engineering,Liaoning University of Technology,Jinzhou 121001,China)

机构地区：[1]辽宁工业大学汽车与交通工程学院,辽宁锦州121001

出　　处：《现代车用动力》2024年第1期1-6,60,共7页Modern Vehicle Power

基　　金：2022年辽宁省教育厅项目(LJKMZ20220978)。

摘　　要：环境感知是无人驾驶汽车的重要基础,目标检测是环境感知的重要一环,随着深度学习的发展,基于深度学习的目标检测算法的精度和速度也在不断提升,逐渐成为主流算法。总结梳理了基于深度学习的目标检测算法以及其最新研究进展,并将其分为基于图像、点云、融合和变形器(Transformer)四大类,分别介绍了其经典算法。最后总结了当前目标检测领域里存在的不足,对目标检测算法的发展进行了展望,为该领域的研究工作提供一些思路。Environmental perception is an important foundation for autonomous vehicles,and object detection is an important part of environmental perception.With the development of deep learning,the accuracy and speed of object detection algorithms based on deep learning are constantly improved,gradually becoming the mainstream algorithm.It summarizes and sorts out deep learning based object detection algorithms and their latest research progress,and divides them into four categories:image based,point cloud based,fusion based and Transformer based.Their classic algorithms are respectively introduced.Finally,the current problems in the field of object detection are summarized,and the future development trend is prospected.They provide valuable ideas for future research in this field.

关 键 词：无人驾驶汽车 环境感知 卷积神经网络 

分 类 号：U463.6[机械工程—车辆工程] TP391.41[交通运输工程—载运工具运用工程]