首页 > 论文 > 光学学报 > 38卷 > 6期(pp:615004--1)

基于改进Faster R-CNN的空中目标检测

Aerial Target Detection Based on Improved Faster R-CNN

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

相比传统图像目标检测算法,基于大数据和深度学习的检测算法无须人工设计特征,且检测性能更稳健。在防空应用背景下,自建了空中目标静态和视频图像数据集用于训练和测试,改进了基于深度学习的目标检测框架Faster R-CNN,将其专用于空中目标检测。结合空中目标检测任务的特点和需求,提出膨胀积累、区域放大、局部标注、自适应阈值、时空上下文等改进策略,弥补了Faster R-CNN对弱小目标和被遮挡目标不敏感的缺陷,提高了检测速度和精度。实验表明,改进后的Faster R-CNN在应对弱小目标、多目标、杂乱背景、光照变化、模糊、大面积遮挡等检测难度较大的情况时,均能获得很好的效果。数据集上测试结果的平局准确率均值较改进之前提高了16.7%,检测速度提高了3倍。

Abstract

Compared with the traditional detectors, the detectors based on large data and deep learning do not require manually designed features and are more robust. Under the background of air defense, we build the images and videos dataset of aerial target for training and test, improve the deep learning-based detector Faster R-CNN, and specialize it in aerial target detection. Aiming at the peculiarities and requirements of aerial target detection, we propose the strategies such as accumulation of dilation, regional amplification, local tagging, adaptive threshold and spatio-temporal context to make up the shortage of Faster R-CNN that small weak or occluded targets can not be detected and improve the detection speed and accuracy. Experimental results show that the improved Faster R-CNN performs well under circumstances such as small weak or multiple targets, clutter, illumination changes, blur and large-area occlusion. Compared to the original Faster R-CNN, the mean average precision is improved by 16.7% on the built dataset, and the speed is 3 times faster.

Newport宣传-MKS新实验室计划
补充资料

中图分类号:TP391.41

DOI:10.3788/aos201838.0615004

所属栏目:机器视觉

基金项目:国防科技预研项目 (40405070102)

收稿日期:2017-12-06

修改稿日期:2018-01-22

网络出版日期:--

作者单位    点击查看

冯小雨:陆军工程大学石家庄校区电子与光学工程系, 河北 石家庄 050003
梅卫:陆军工程大学石家庄校区电子与光学工程系, 河北 石家庄 050003
胡大帅:陆军工程大学石家庄校区电子与光学工程系, 河北 石家庄 050003

联系人作者:冯小雨(826782445@qq.com)

备注:冯小雨(1993-),男,硕士研究生,主要从事深度学习与计算机视觉方面的研究。E-mail: 826782445@qq.com

【1】Wang Y X, Shi D L, Yang N. Fusion algorithm of frame difference and mean shift for high-speed UAVs’ tracking[J]. Journal of Naval Aeronautical and Astronautical University, 2016, 31(4): 437-441.
王宜贤, 石德乐, 杨宁. 帧差法和Mean shift算法融合的高速无人机目标跟踪[J]. 海军航空工程学院学报, 2016, 31(4): 437-441.

【2】Chu J. Detection of moving target in dynamic background based on optical flow[J]. Journal of Nanchang University of Aeronautics and Astronautics, 2011, 25(3): 1-6.
储珺. 基于光流的动态背景运动目标检测算法[J]. 南昌航空大学学报, 2011, 25(3): 1-6.

【3】Li S S, Zhao G P, Wang J Y. Distractor-aware object tracking based on multi-feature[J]. Acta Optica Sinica, 2017, 37(5): 0515005.
李双双, 赵高鹏, 王建宇. 基于特征融合和尺度自适应的干扰感知目标跟踪[J]. 光学学报, 2017, 37(5): 0515005.

【4】Tan Y, Wang Y J. On a combination of background difference improved CamShift target tracking method[J]. Journal of Southwest China Normal University (Natural Science Edition), 2016, 41(9): 120-125.
谭艳, 王宇俊. 一种结合背景差分的改进CamShift目标跟踪方法[J]. 西南师范大学学报(自然科学版), 2016, 41(9): 120-125.

【5】Yin L, Gao K, Bai T Z. Research on oilcan target segmentation in remote sensing image based on improved Otsu algorithm[J]. Optical Technique, 2012, 38(2): 197-201.
殷亮, 高昆, 白廷柱. 基于改进Otsu法的遥感图像油罐目标分割研究[J]. 光学技术, 2012, 38(2): 197-201.

【6】Yu K, Jia L, Chen Y Q, et al. Deep learning: yesterday, today,and tomorrow[J]. Journal of Computer Research and Development, 2013, 50(9): 1799-1804.
于凯, 贾磊, 陈宇强, 等. 深度学习的昨天, 今天和明天[J].计算机研究与发展, 2013, 50(9): 1799-1804.

【7】Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554.

【8】Lecun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444.

【9】Ren S Q, He K M, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]∥Proceedings of Advances in Neural Information Processing Systems, 2015: 91-99.

【10】Liu W, Anguelov D, Erhan D, et al. Ssd: single shot multibox detector[C]. European Conference on Computer Vision, 2016: 21-37.

【11】Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2017: 6517-6525.

【12】Hare S, Golodetz S, Saffari A, et al. Struck: structured output tracking with kernels[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(10): 2096-2109.

【13】Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]∥Proceedings of 2014 IEEE International Conference on Computer Vision and Pattern Recognition, 2014: 580-587.

【14】He K M, Zhang X Y, Ren S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37 (9): 1904-1916.

【15】Girshick R. Fast R-CNN[C]∥Proceedings of 2015 IEEE International Conference on Computer Vision, 2015: 1440-1448.

【16】Qu S R, Yang H H. Multi-target detection and tracking of video sequence based on Kalman_BP neural network[J]. Infrared and Laser Engineering, 2013, 42(9): 2553-2560.
曲仕茹, 杨红红. 采用Kalman_BP神经网络的视频序列多目标检测与跟踪[J]. 红外与激光工程, 2013, 42(9): 2553-2560.

【17】Li S J, Zhu Z F. Track-before-detect algorithm via particle filtering[J]. Infrared and Laser Engineering, 2009, 38(2): 352-357.
李少军, 朱振福. 采用粒子滤波的先跟踪后检测算法[J]. 红外与激光工程, 2009, 38(2): 352-357.

【18】Jia Y, Shelhamer E, Donahue J, et al. Caffe: convolutional architecture for fast feature embedding[C]∥Proceedings of the 22nd ACM International Conference on Multimedia, 2014: 675-678.

【19】Wang D F, Wang Y D, Wang J W. Moving target detection algorithm based on improved Gaussian mixture model[J]. Laser Technology, 2014, 38(6): 776-779.
王东方, 王玉德, 王景武. 基于改进的混合高斯模型的运动目标检测方法[J]. 激光技术, 2014, 38(6): 776-779.

【20】Yang C C, Gu G H, Qian W X, et al. Hardware implementation of infrared image registration based on the Harris corner of KLT tracking[J]. Infrared Technology, 2013, 35(10): 632-637.
杨陈晨, 顾国华, 钱惟贤, 等. 基于Harris角点的KLT跟踪红外图像配准的硬件实现[J]. 红外技术, 2013, 35(10): 632-637.

【21】Jiang Q Y, Li P, Sun L. Application of Otsu method in motion detection system[J]. Journal of Computer Applications, 2011, 31(1): 260-262.
江亲瑜, 李平, 孙兰. 最大类间方差算法在运动检测系统中的应用[J]. 计算机应用, 2011, 31(1): 260-262.

【22】Cui Z, Xiao S, Feng J, et al. Recurrently target-attending tracking[C]∥Proceedings of 2016 IEEE International Conference on Computer Vision, 2016: 1449-1458.

【23】Cai Y Z, Yang D D, Mao N, et al. Visual tracking algorithm based on adaptive convolutional features[J]. Acta Optica Sinica, 2017, 37(3): 0315002.
蔡玉柱, 杨德东, 毛宁, 等. 基于自适应卷积特征的目标跟踪算法[J]. 光学学报, 2017, 37(3): 0315002.

引用该论文

Feng Xiaoyu,Mei Wei,Hu Dashuai. Aerial Target Detection Based on Improved Faster R-CNN[J]. Acta Optica Sinica, 2018, 38(6): 0615004

冯小雨,梅卫,胡大帅. 基于改进Faster R-CNN的空中目标检测[J]. 光学学报, 2018, 38(6): 0615004

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF