基于深度学习的小目标检测算法综述  

Review on Small Target Detection Based on Deep Learning

作　　者：张新 郭福亮[1] 梁英杰 陈修亮 ZHANG Xin;GUO Fu-liang;LIANG Ying-jie;CHEN Xiu-liang(Department of Computer and Data Engineering,Naval University of Engineering;Nation Key Laboratory of Science and Technology on Vessel Integrated Power System,Naval University of Engineering,Wuhan 430033,China;Information Center of Training Management Department of Military Commission,Beijing 100097,China)

机构地区：[1]海军工程大学计算机与数据工程教研室 [2]海军工程大学舰船综合电力技术国防科技重点实验室,湖北武汉430033 [3]中央军委训练管理部信息中心,北京100097

出　　处：《软件导刊》2020年第5期276-280,共5页Software Guide

基　　金：国家自然科学基金项目(61701517)。

摘　　要：由于小目标分辨率低、携带的信息少,现有基于深度学习的目标检测算法对小目标检测精度远远小于对大目标的检测精度。针对小目标检测精度问题,以基于深度学习的目标检测为切入点,系统总结了基于深度学习的目标检测代表算法,并以检测精度和检测速度为标准分析各种算法优缺点。将能有效提高小目标检测精度的方法进行分类汇总,介绍了各种方法的相关应用。针对目标检测及小目标检测存在的问题,对未来目标检测领域发展趋势进行了预测与展望。Due to the low resolution of small target and the lack of information,the existing target detection algorithm based on deep learning has much less detection accuracy for small target than for large target.Aiming at the problem of small target detection,the rep⁃resentative algorithms of target detection based on deep learning are systematically summarized and introduced,and the advantages and disadvantages of various algorithms are analyzed according to the detection accuracy and detection speed.Then,with small target detection as the main method,the commonly used improved methods which can effectively improve the accuracy of small target detec⁃tion are classified and summarized,and the related applications of various methods are introduced.Finally,aiming at the problems of target detection and small target detection,the future development trend of target detection is predicted and prospected.

关 键 词：目标检测 小目标 深度学习 RCNN SSD YOLO 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]