基于YOLO改进算法的远程塔台运动目标检测
|
摘要
远程塔台由于其低成本高时效远程实时控制技术正越来越受到民航业界的青睐,其中运动目标自动检测和显示是远程塔台的核心技术,作为增强现实技术更好地为管制员提供服务。在分析远程塔台机场场面背景复杂、场面目标多为远场景、小目标等特点基础上,提出了一种改进的You Only Look Once(YOLO)算法来实现远程塔台运动目标的检测,算法核心思想以Darknet-53为基础网络,多尺度预测边界框,以运动目标图像坐标的偏移量作为边框长宽的线性变换来实现边框的回归,改善了传统YOLO算法损失函数不同大小的边框未做区分的问题,提高了检测准确性和速度。机场真实数据实验表明,该算法能快速、准确的检测出远程塔台的运动目标,并准确的回归运动目标边框及分类。
Remote tower is becoming more and more popular in civil aviation industry because of its low cost,high efficiency and remote real-time control technology. The automatic detection and display of moving targets is the core technology of the remote tower,and it serves as a better augmented reality technology for controllers. The remote tower was analyzed,which was characterized by complex airport scenes,scene targets such as far scenes,small targets,etc. Therefore,An improved You Only Look Once( YOLO) algorithm did proposed to detect remote moving targets. The algorithm was based on the Darknet-53 network. The algorithm predicted bounding boxes from multiple scales. The offset of the moving target image coordinate was used as the frame length and width. The regression of the bounding box was achieved by the linear transformation of the offset. The loss function of the traditional YOLO algorithm does not distinguish the size of the bounding box. The above problem was solved by the algorithm. And the accuracy of the detection was higher and faster by using this algorithm. Experiments on real airport data show that the algorithm can detect the moving targets of long-distance tower quickly and accurately,and accurately regress the moving object boundaries and classification.
Remote tower is becoming more and more popular in civil aviation industry because of its low cost,high efficiency and remote real-time control technology. The automatic detection and display of moving targets is the core technology of the remote tower,and it serves as a better augmented reality technology for controllers. The remote tower was analyzed,which was characterized by complex airport scenes,scene targets such as far scenes,small targets,etc. Therefore,An improved You Only Look Once( YOLO) algorithm did proposed to detect remote moving targets. The algorithm was based on the Darknet-53 network. The algorithm predicted bounding boxes from multiple scales. The offset of the moving target image coordinate was used as the frame length and width. The regression of the bounding box was achieved by the linear transformation of the offset. The loss function of the traditional YOLO algorithm does not distinguish the size of the bounding box. The above problem was solved by the algorithm. And the accuracy of the detection was higher and faster by using this algorithm. Experiments on real airport data show that the algorithm can detect the moving targets of long-distance tower quickly and accurately,and accurately regress the moving object boundaries and classification.
引文
1陈国玉.序列图像中运动目标检测算法研究[D].大连:大连理工大学,2010Chen Guoyu.Moving object detection algorithm in sequence images[D].Dalian:Dalian University of Technology,2010
2万缨,韩毅,卢汉清.运动目标检测算法的探讨[J].计算机仿真,2006,23(10):221-226Wan Ying,Han Yi,Lu Hanqing.Research on moving target detection algorithm[J].Computer Simulation,2006,23(10):221-226
3王晓卫,宁固.一种改进的基于光流的运动目标的检测算法[J].武汉大学学报(信息科学版),2003,28(3):351-153Wang Xiaowei,Ning Gu.An improved moving object detection algorithm based on optical flow[J].Journal of Wuhan University(Information Science Edition),2003,28(3):351-353
4 Zhan C,Duan X,Xu S,et al.An improved moving object detection algorithm based on frame difference and edge detection[C]//Fourth International Conference on Image and Graphics(ICIG).Beijing:IEEE,2007:519-523
5王孝艳,张艳珠,董慧颖,等.运动目标检测的三帧差法算法研究[J].沈阳理工大学学报,2011,30(6):82-85Wang Xiaoyan,Zhang Yanzhu,Dong Huiying,et al.Three-frame difference algorithm for moving object detection[J].Journal of Shenyang University of Technology,2011,30(6):82-85
6阮涛涛,姚明海,瞿心昱.基于视觉的人体运动分析综述[J].计算机系统应用,2011,20(2):245-254Ruan Taotao,Yao Minghai,Qu Xinyu.A review of vision-based human motion analysis[J].Computer Systems Applications,2011,20(2):245-254
7李文斌,周晓敏,王长松.一种基于背景减法的运动目标检测算法[J].工程科学学报,2008,30(2):212-216Li Wenbin,Zhou Xiaomin,Wang Changsong.A moving object detection algorithm based on background subtraction[J].Journal of Engineering Sciences,2008,30(2):212-216
8周同雪,朱明.视频图像中的运动目标检测[J].液晶与显示,2017,32(1):40-47Zhou Tongxue,Zhu Ming.Moving object detection in video images[J].LCD and display,2017,32(1):40-47
9 Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Conference on Computer Vision and Pattern Recognition(CVPR).Columbus:IEEE,2014:580-587
10 He K,Zhang X,Ren S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,37(9):1904-1916
11 Girshick R.Fast R-CNN[C]//International Conference on Computer Vision(ICCV).Santiago,2015:1440-1448
12 Ren S,He K,Girshick R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,39(6):1137-1149
13 Dai J,Li Y,He K,et al.R-FCN:object detection via region-based fully convolutional networks[J].2016,26(7):235-238
14 Redmon J,Divvala S,Girshick R,et al.You Only Look Once:unified,real-time object detection[C]//Conference on Computer Vision and Pattern Recognition(CVPR).San Francisco:IEEE,2016:779-788
15 Liu W,Anguelov D,Erhan D,et al.SSD:Single Shot MultiBox Detector[C]//European Conference on Computer Vision(ECCV).San Francisco,CA,USA:IEEE Conference,2016:6517-6525
16 Redmon J,Divvala S,Girshick R,et al.You Only Look Once:unified,real-time object detection[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,14(6):779-788
17 Redmon J,Farhadi A.Yolo 9000:Better,faster,stronger[C]//Computer Vision and Pattern Recognition(CVPR).San Francisco:IEEE,2017:6517-6525
18 He K M,Zhang X Y,Ren S Q.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.San Francisco:IEEE,2015:770-778
2万缨,韩毅,卢汉清.运动目标检测算法的探讨[J].计算机仿真,2006,23(10):221-226Wan Ying,Han Yi,Lu Hanqing.Research on moving target detection algorithm[J].Computer Simulation,2006,23(10):221-226
3王晓卫,宁固.一种改进的基于光流的运动目标的检测算法[J].武汉大学学报(信息科学版),2003,28(3):351-153Wang Xiaowei,Ning Gu.An improved moving object detection algorithm based on optical flow[J].Journal of Wuhan University(Information Science Edition),2003,28(3):351-353
4 Zhan C,Duan X,Xu S,et al.An improved moving object detection algorithm based on frame difference and edge detection[C]//Fourth International Conference on Image and Graphics(ICIG).Beijing:IEEE,2007:519-523
5王孝艳,张艳珠,董慧颖,等.运动目标检测的三帧差法算法研究[J].沈阳理工大学学报,2011,30(6):82-85Wang Xiaoyan,Zhang Yanzhu,Dong Huiying,et al.Three-frame difference algorithm for moving object detection[J].Journal of Shenyang University of Technology,2011,30(6):82-85
6阮涛涛,姚明海,瞿心昱.基于视觉的人体运动分析综述[J].计算机系统应用,2011,20(2):245-254Ruan Taotao,Yao Minghai,Qu Xinyu.A review of vision-based human motion analysis[J].Computer Systems Applications,2011,20(2):245-254
7李文斌,周晓敏,王长松.一种基于背景减法的运动目标检测算法[J].工程科学学报,2008,30(2):212-216Li Wenbin,Zhou Xiaomin,Wang Changsong.A moving object detection algorithm based on background subtraction[J].Journal of Engineering Sciences,2008,30(2):212-216
8周同雪,朱明.视频图像中的运动目标检测[J].液晶与显示,2017,32(1):40-47Zhou Tongxue,Zhu Ming.Moving object detection in video images[J].LCD and display,2017,32(1):40-47
9 Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Conference on Computer Vision and Pattern Recognition(CVPR).Columbus:IEEE,2014:580-587
10 He K,Zhang X,Ren S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,37(9):1904-1916
11 Girshick R.Fast R-CNN[C]//International Conference on Computer Vision(ICCV).Santiago,2015:1440-1448
12 Ren S,He K,Girshick R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,39(6):1137-1149
13 Dai J,Li Y,He K,et al.R-FCN:object detection via region-based fully convolutional networks[J].2016,26(7):235-238
14 Redmon J,Divvala S,Girshick R,et al.You Only Look Once:unified,real-time object detection[C]//Conference on Computer Vision and Pattern Recognition(CVPR).San Francisco:IEEE,2016:779-788
15 Liu W,Anguelov D,Erhan D,et al.SSD:Single Shot MultiBox Detector[C]//European Conference on Computer Vision(ECCV).San Francisco,CA,USA:IEEE Conference,2016:6517-6525
16 Redmon J,Divvala S,Girshick R,et al.You Only Look Once:unified,real-time object detection[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,14(6):779-788
17 Redmon J,Farhadi A.Yolo 9000:Better,faster,stronger[C]//Computer Vision and Pattern Recognition(CVPR).San Francisco:IEEE,2017:6517-6525
18 He K M,Zhang X Y,Ren S Q.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.San Francisco:IEEE,2015:770-778
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |