深度学习的目标检测典型算法及其应用现状分析  

Analysis of typical target detection algorithm based on deep learning and its application status

作　　者：侯学良[1] 单腾飞 薛靖国 Hou Xueliang;Shan Tengfei;Xue Jingguo(School of Economics and Management,North China Electric Power University,Beijing 102206,China)

机构地区：[1]华北电力大学经济与与管理学院,北京102206

出　　处：《国外电子测量技术》2022年第6期165-174,共10页Foreign Electronic Measurement Technology

基　　金：国家自然科学基金(71171081);北京市自然科学基金(9162014)项目资助。

摘　　要：目标检测是利用图像处理技术对输入图像中的兴趣目标进行分类和定位。深度学习凭借强大的表征和建模能力,使得目标检测的效率大大提升。首先回顾了传统目标检测方法的检测过程以及存在的问题;然后,分别从两阶段和单阶段两大方面,对基于深度学习的典型目标检测算法进行了比较,介绍了目标检测算法常用的性能评价指标和数据集。在此基础上,总结了当前目标检测算法的应用领域,分析了目标检测研究中需要进一步深入探究的问题,并对未来目标检测的发展趋势给出了相关建议。Target detection is to classify and locate the interested targets in the input image by using image processing technology.With its powerful representation and modeling ability,deep learning greatly improves the efficiency of target detection.Firstly,this paper reviews the detection process and existing problems of traditional target detection methods;Then,the typical target detection algorithms based on deep learning are compared from two-stage and single-stage aspects,and the commonly used performance evaluation indexes and data sets of target detection algorithms are introduced.On this basis,this paper summarizes the application fields of current target detection algorithms,analyzes the problems that need to be further explored in the research of target detection,and gives relevant suggestions on the development trend of target detection in the future.

关 键 词：目标检测 深度学习 检测模型 计算机视觉 

分 类 号：TP38[自动化与计算机技术—计算机系统结构]