改进YOLO v2的装甲车辆目标识别

详细信息    查看全文 | 推荐本文 |

英文篇名：Improved YOLO v2 for Target Recognition of Armored Vehicles 作者：王曙光 ; 吕攀飞 英文作者：WANG Shu-guang;LYU Pan-fei;Department of Ordnance Engineering,Army Artillery and Air Defense Forces Academy of PLA; 关键词：装甲目标识别 ; 维度聚类 ; YOLO ; v2 ; anchor 英文关键词：armored target recognition;;dimensional clustering;;YOLO v2;;anchor 中文刊名：JYXH 英文刊名：Computer and Modernization 机构：中国人民解放军陆军炮兵防空兵学院兵器工程系; 出版日期：2018-09-15 出版单位：计算机与现代化 年：2018 期：No.277 语种：中文; 页：JYXH201809016 页数：5 CN：09 ISSN：36-1137/TP 分类号：72-75+83

摘要

军事目标识别技术是军事信息处理的一个重要内容,对于实现军事装备信息化、智能化起着不可忽视的作用。近年来随着深度卷积神经网络在图像识别领域的广泛应用,各种基于图像目标识别任务的网络结构层出不穷,因此将这项新技术应用于军事目标的识别具有极强的现实意义和军事应用价值。本文以目前具有最佳识别效果的YOLO v2网络为基础,通过维度聚类重新确定最优的anchor个数及其宽高维度,并制作以明显特征为目标区域的装甲车辆数据集,使得该网络对装甲目标的识别更为精确。通过实验验证,该方法能有效地对特定装甲目标进行实时精确识别。
        The technology of military target recognition is an important part of military information processing,which plays an important role in realizing the informatization and intelligentization of military equipment. In recent years,with the wide application of convolutional neural network in image recognition field,a variety of network structures based on image recognition task emerge in an endless stream. So it is of great practical significance and military application value to apply the new technology in military target recognition. Based on the YOLO v2 network which has the best recognition effect at present,this paper redefines the optimal number of anchors and their width and height dimensions by dimension clustering,and makes the armored vehicle data set with obvious features as the target area,so that the network can recognize the armored targets more accurately. Experimental results show that the method can effectively identify the specific armored targets in real time.

引文

[1]Le Cun Y,Bengio Y,Hinton G.Deep learning[J].Nature,2015,521(7553):436-444.
    [2]Girshick R.Fast R-CNN[C]//Proceedings of the 2015IEEE International Conference on Computer Vision.2015:1440-1448.
    [3]Ren Shaoqing,He Kaiming,Girshick R,et al.Faster RCNN:Towards real-time object detection with region proposal networks[C]//Proceedings of the 2015 Advances in Neural Information Processing Systems.2015:91-99.
    [4]Redmon J,Divvala S,Girshick R,et al.You only look once:Unified,real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition.2016:779-788.
    [5]Liu Wei,Anguelov D,Erhan D,et al.SSD:Single shot multibox detector[C]//Proceedings of the 2016 European Conference on Computer Vision.2016:21-37.
    [6]Szegedy C,Liu Wei,Jia Yangqing,et al.Going deeper with convolutions[C]//Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition.2015:1-9.
    [7]张素洁,赵怀慈.最优聚类个数和初始聚类中心点选取算法研究[J].计算机应用研究,2017,34(6):1617-1620.
    [8]吕欣然,陈洁,张立保,等.基于显著特征聚类的遥感图像感兴趣区域检测[J].光学学报,2015,35(S1):103-108.
    [9]Redmon J,Farhadi A.YOLO9000:Better,Faster,Stronger[DB/OL].https://arxiv.org/pdf/1612.08242v1.pdf,2016-12-25.
    [10]Kong Tao,Yao Anbang,Chen Yurong,et al.Hyper Net:Towards Accurate Region Proposal Generation and Joint Object Detection[DB/OL].https://arxiv.org/pdf/1604.00600v1.pdf,2016-04-03.
    [11]孙皓泽,常天庆,王全东,等.一种基于分层多尺度卷积特征提取的坦克装甲目标图像检测方法[J].兵工学报,2017,38(9):1681-1691.
    [12]Russakovsky O,Deng Jia,Su Hao,et al.Image Net large scale visual recognition challenge[J].International Journal of Computer Vision,2015,115(3):211-252.
    [13]Lim T R,Guntoro A T.Car recognition using Gabor filter feature extraction[C]//Proceedings of the 2002 Asia-Pacific Conference on Circuits and Systems.2002,2:451-455.
    [14]Hinton G,Deng Li,Yu Dong,et al.Deep neural networks for acoustic modeling in speech recognition:The shared views of four research groups[J].IEEE Signal Processing Magazine,2012,29(6):82-97.
    [15]Zeiler M D,Fergus R.Visualizing and understanding convolutional networks[C]//Proceedings of the 2014 European Conference on Computer Vision.2014:818-833.
    [16]魏湧明,全吉成,侯宇青阳.基于YOLO v2的无人机航拍图像定位研究[J].激光与光电子学进展,2017,54(11):101-110.