基于深度学习的目标检测算法综述  

Study on Improved Ant Colony Path Planning Algorithm

作　　者：邹军 张世义[2] 李军[1] ZOU Jun;ZHANG Shiyi;LI Jun(School of Mechanical,Electrical and Vehicle Engineering,Chongqing Jiao tong University,Chongqing 400074,China;School of Shipping and Naval Architecture,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学机电与车辆工程学院,重庆400074 [2]重庆交通大学航海与船舶工程学院,重庆400041

出　　处：《传感器世界》2023年第8期9-15,I0002,共8页Sensor World

基　　金：重庆市研究生联合培养基地项目(No.JDLHPYJD2018003)。

摘　　要：基于深度学习的目标检测算法近年来倍受关注。文章首先介绍了传统目标检测器、基本性能指标和主流数据集;然后,按照基于回归的一阶段和基于候选框的两阶段两个大类分别阐述深度学习为基础的目标检测算法,对算法的流程步骤、网络结构深入分析,梳理算法发展中的常规改进操作和创新点,对比分析它们的优势劣势以及导致这些结果的原因;最后,总结目标检测算法的发展历程和展望未来该领域的研究趋势,提出该领域存在的问题和未来的改进方向。Object detection algorithms based on deep learning have attracted much attention in recent years.Firstly,traditional target detectors,basic performance indicators,and mainstream datasets are introduced.Then,according to the two major categories of regressionbased one-stage and candidate box-based two-stage respectively,the deep learning-based object detection algorithm is expounded,the process steps and network structure of the algorithm are analyzed in depth,the conventional improvement operations and innovation points in the development of the algorithm are sorted out,and their advantages and disadvantages and the reasons for these results are compared and analyzed.Finally,the development process of object detection algorithm is summarized,the future research trend in this field is prospected,and the existing problems in this field and the future improvement direction are proposed.

关 键 词：目标检测算法 一阶段、两阶段 特征提取 深度学习 

分 类 号：TN911.73[电子电信—通信与信息系统]