计算机科学 ›› 2023, Vol. 50 ›› Issue (11A): 220900050-8.doi: 10.11896/jsjkx.220900050
路琪1, 于元强1, 许道明1, 张琦2
LU Qi1, YU Yuanqiang1, XU Daoming1, ZHANG Qi 2
摘要: 低空慢速小型目标检测一直是预警探测领域关注的重点和难点。目前,基于神经网络的主流目标检测算法在设计时主要考虑应用于VOC数据集或COCO数据集,在特定场景下检测精度不够理想。针对复杂背景下小型旋翼无人机目标检测的特定检测场景,提出一种基于改进YOLOv5的小型旋翼无人机目标检测算法。首先,增加小目标检测层以获得大尺寸的浅层特征图,从而提升算法对小目标的检测能力;其次,针对小型旋翼无人机尺寸不一的问题,利用K-Means++聚类算法对先验框的尺寸进行优化并将其与各特征层进行匹配;最后,使用Mosaic-SOD方法进行数据增强以及改进损失函数,增强算法对小目标的感知能力以及提高网络训练效率。将改进后的算法应用在复杂背景下的小型旋翼无人机目标检测中,实验结果表明,相较于原始YOLOv5算法,该算法在小型旋翼无人机目标检测上具有更高的检测精度和特征提取能力,虽然检测速度有一定下降,但通过对可见光视频流进行检测可知其仍能够满足实时性的要求。
中图分类号:
[1]GRANT R.RPAs FOR ALL[J].Air Force Magazine,2012,95(8):54-57 [2]XIE X.US Army releases anti-drone technology manual (Part 1)[J].Modern Military,2017(7):91-100. [3]XIE X.US Army releases anti-drone technology manual (Part 2)[J].Modern Military,2017(8):85-89. [4]GUELFI E A,BUDDHIKA J,ROBISON T.The Imperative for the U.S.Military to Develop a Counter-UAS Strategy[J].JFQ:Joint Force Quarterly,2020(97):4-12. [5]MA W,CHIGAN X X.Research on Development of Anti-UAV Technology[J].Aero Weaponry,2020,27(6):19-24. [6]JIE C,MIAO Z,YE T T.Research on the development of anti-UAV systems in U.S.military service[J].Airborne Missile,2020(12):36-42. [7]CHEN W.A Survey on Low-Slow-Small UAV Countermeasure Equipment andTheoretical[J].Henan Science and Technology,2022,41(2):10-13. [8]ZHU M Z,CHEN X,LIU X,et al.Situation and key technology of tactical laser anti-UAV[J].Infrared and Laser Engineering,2021,50:188-200. [9]WU Y F.Research on LSS-Target(the Low altitude,Slowspeed andSmall target) in complex background [D].Changchun:Changchun university of Chinese Academy of Sciences,China(Chinese Academy of Sciences Institute of Optical Precision Machinery and Physical),2020. [10]XU D M,ZHANG H w.Overview of Radar LSS Target Detection Technology[J].Modern Defence Technology,2018,46(1):148-155. [11]MAO T.Research on Small Object Detection Algorithm Based onYOLOv5[D].Anhui University of Science and Technology,2021. [12]REN J,WANG Z J,ZHANG Y F,et al.YOLOv5-R:lightweight real-time detection based on improved YOLOv5[J].Journal of Electronic Imaging,2022,31(3):033033. [13]XIANG X Z,WANG Z Y,QIAO Y L.An Improved YOLOv5Crack Detection Method Combined with Transformer[J].IEEE Sensors Journal,2022,22(14):1 [14]JIANG X K,HU H C,LIU X,et al.A smoking behavior detection method based on the YOLOv5 network[J].Journal of Physics:Conference Series,2022,2232:012001. [15]DONG X D,YAN S,DUAN C Q.A lightweight vehicles detection network model based on YOLOv5[J].Engineering Applications of Artificial Intelligence,2022,113:104914. |
|