[1]沈旭,孟巍,程小辉,等.机载平台下基于深度检测网络的目标跟踪重捕算法[J].红外技术,2020,42(7):624-631.[doi:10.11846/j.issn.1001_8891.202007004]
 SHEN Xu,MENG Wei,CHENG Xiaohui,et al.Object Tracking and Recapture Model Based on Deep Detection Network Under Airborne Platform[J].Infrared Technology,2020,42(7):624-631.[doi:10.11846/j.issn.1001_8891.202007004]
点击复制

机载平台下基于深度检测网络的目标跟踪重捕算法
分享到:

《红外技术》[ISSN:1001-8891/CN:CN 53-1053/TN]

卷:
42卷
期数:
2020年第7期
页码:
624-631
栏目:
出版日期:
2020-07-23

文章信息/Info

Title:
Object Tracking and Recapture Model Based on Deep Detection Network Under Airborne Platform

文章编号:
1001-8891(2020)07-0624-08
作者:
沈旭1孟巍2程小辉3王新政3
1. 岭南师范学院 信息工程学院;
2. 山东电力科学研究院;
3. 桂林理工大学 信息科学与工程学院

Author(s):
SHEN Xu1MENG Wei2CHENG Xiaohui3WANG Xinzheng3
1. School of Information Engineering, Lingnan Normal University;
2. Shandong Electric Power Research Institute;
3. College of Information Science and Engineering, Guilin University of Technology

关键词:
目标跟踪深度学习Siamese网络轮廓模板目标检测
Keywords:
object tracking deep learning siamese network contour extraction network object detection
分类号:
TN219, TP181, TP391
DOI:
10.11846/j.issn.1001_8891.202007004
文献标志码:
A
摘要:
目标检测与跟踪是机载光电设备至关重要的功能模块,其检测跟踪的性能直接关系到目标感知的精度。近年来基于Siamese网络的改进跟踪算法在各种挑战性的数据集上取得了优异的效果,但大多数改进算法采用局部搜索策略,无法更新模板,且模板会引入背景干扰,最终因跟踪点漂移导致跟踪失败。为了解决这些问题,本文提出了一种结合目标边缘检测的改进全连接Siamese跟踪算法,该算法利用目标的轮廓模板代替边界框模板,减少了背景杂波的干扰;同时,在Siamese网络的基础上增加了一路改进tiny-YOLOv3目标检测网络,利用K均值聚类找到最合适的锚框(anchor box),引入了扩张模块层来扩展感受野,增加了系统的抗遮挡能力,提高机载光电设备的目标捕获概率。在基准测试数据集以及挂飞数据集基础上的仿真测试性能表明本文提出的改进模型特别适合机载光电设备在跟踪与重捕复杂环境下的运动目标,在长期跟踪中能够更好地适应目标的变形和遮挡,提升系统响应时间与适应性。
Abstract:
Object detection and tracking is an essential module in airborne optoelectronic equipment, and its performance is directly related to the accuracy of object perception. Improved Siamese network tracking algorithms have produced excellent results for various challenging datasets recently, but most of the improved algorithms use local fixed search strategies, which cannot update the template. In addition, the template will introduce background interference, which will result in tracking drift and eventually cause tracking failure. To solve these problems, this paper proposes an improved fully connected Siamese tracking algorithm combined with object contour extraction and object detection; the algorithm uses the contour template of the target instead of the bounding box template to reduce the background clutter interference. A branch is added to the Siamese network to improve the tiny-YOLOv3 object detection network, where K-means clustering is used to find the most suitable anchor box. An expansion module layer is introduced to expand the receptive field. Therefore, our proposed model increases the anti-occlusion ability of the system and improves the object recapture probability of airborne optoelectronic equipment. The results of a simulation of benchmark test data set and a flight dataset show that the improved model is especially suitable for tracking and recapture of moving objects in complex environments; in addition, it can better adapt to deformed or occluded objects in long-term tracking, which improves the system response time and adaptability.

参考文献/References:

[1] 王爽, 邵艳明, 宋敏敏, 等. 一种面向目标跟踪性能的红外图像自适应增强方法[J]. 红外技术, 2019, 41(7): 646-653.
WANG S, SHAO Y M, SONG M M, et al. An adaptive enhancement method for infrared image based on target tracking performance[J]. Infrared Technology, 2019, 41(7) : 646-653.
[2] KONG L, HUANG D, QIN J, et al. A Joint Framework for Athlete Tracking and Action Recognition in Sports Videos[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 23(5): 1-21.
[3] YANG H, WEN J, WU X J, et al. An Efficient Edge Artificial Intelligence Multi-pedestrian Tracking Method with Rank Constraint[J]. IEEE Transactions on Industrial Informatics, 2019, 12(34): 1-1.
[4] MA Chao, YANG M, ZHAN C M, et al. Hierarchical features for visual tracking[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017, 48(9): 21-31.
[5] 齐永锋, 王梦媛. 基于LBP与核相关滤波器的运动目标跟踪算法[J]. 红外技术, 2019, 41(6): 572-576.?
QI Yongfeng, WANG Mengyuan. Moving Target Tracking Algorithm Based on LBP and Kernel Correlation Filter[J]. Infrared Technology, 2019, 41(6): 572-576.
[6] Poppenk J, Norman K A . Multiple-object Tracking as a Tool for Parametrically Modulating Memory Reactivation[J]. Journal of Cognitive Neuroscience, 2017: 1-15.
[7] Galteri L, Seidenari L, Bertini M, et al. Spatio-Temporal Closed-Loop Object Detection[J]. IEEE Transactions on Image Processing, 2017, 26(3): 1253-1263.
[8] WANG N, YEUNG D Y. Learning a Deep Compact Image Representation for Visual Tracking[C]//Advances in Neural Information Processing Systems, 2013, 38(18): 5045-5068.
[9] WANG L, OUYANG Wanli, WANG Xiaogang, et al. Visual tracking with fully convolutional networks[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015, 125(1-3): 3-18.
[10] Bertinetto L, Valmadre J, Henriques J F, et al. Fully-Convolutional Siamese Networks for Object Tracking[C]//European Conference on Computer Vision. Springer, 2016: 850-865.
[11] LUO H, XU L, HUI B, et al. Status and prospect of target tracking based on deep learning[J]. Infrared & Laser Engineering, 2017, 46(5): 502002.
[12] 焦安波, 何淼, 罗海波. 一种改进的HED网络及其在边缘检测中的应用[J]. 红外技术, 2019, 41(1): 072-77.
JIAO Anbo, HE Miao, LUO Haibo. Research on Significant Edge Detection of Infrared Image Based on Deep Learning[J]. Infrared Technology, 2019, 41(1): 072-77.
[13] WU Y, LIM J, YANG M H. Online Object Tracking: A Benchmark[C]//Computer Vision and Pattern Recognition (CVPR), 2013: 019-27.
[14] WANG P, CHEN P, YE Y, et al. Understanding Convolution for Semantic Segmentation[C]//IEEE Winter Conference on Applications of Computer Vision, 2018: 1287-1299.
[15] Held D, Thrun S, Savarese S. Learning to Track at 100 FPS with Deep Regression Networks[C]// European Conference on Computer Vision, 2016: 29-37.
[16] Erickson K J, Hanna P M, Westerkamp L A, et al. Evaluation of the VIVID confirmatory identification module[C]//Proc. SPIE, 2007, 6566: 65660B-65660B-12.
[17] Carvalho T, De Rezende E R S, Alves M T P, et al. Exposing computer generated images by eye’s region classification via transfer learning of VGG19 CNN[C]//16th IEEE International Conference on Machine Learning and Applications (ICMLA), 2017: 866-870.
[18] Akiba T, Suzuki S, Fukuda K. Extremely large minibatch SGD: training resnet-50 on imagenet in 15 minutes[J]. arXiv preprint arXiv, 2017, 1711: 04325.
[19] Zeiler M D, Fergus R. Visualizing and understanding convolutional networks BT[J]. Comput. Vis ECCV, 2014: 818-833.
[20] TAO R, Gavves E, Smeulders A W M . Siamese Instance Search for Tracking[C]//Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 1063-6919.?
[21] Redmon J, Farhadi A. YOLOv3: An Incremental Improvement[J]. Computer Vision and Pattern Recognition, 2018, 12(7): 64-78.
[22] WANG Q, GAO J, XING J, et al. DCFnet: Discriminant correlation filters network for visual tracking[J]. Computer Vision and Pattern Recognition, 2017, 1704: 04057.
[23] Valmadre J, Bertinetto L, Henriques J F, et al. End-to-end representation learning for correlation filter based tracking[C]//Proc. IEEE Conf Comput. Vis. Pattern Recognit., 2017: 5000-5008.
[24] LI B, YAN J, WU W, et al. High performance tracking with Siamese region proposal network[C]//Proc. IEEE Comput. Vis.pattern Reconit. (CVPR), 2018: 8971-8980.
[25] Uzkent B, Seo Y W. EnKCF: Ensemble of Kernelized Correlation Filters for High-Speed Object Tracking[C]//IEEE Winter Conference on Applications of Computer Vision, 2018: 77-89.?

相似文献/References:

[1]彭青艳,赵勋杰,陈家波. 跟踪窗口尺寸自适应调整的粒子滤波跟踪算法[J].红外技术,2012,34(10):568.
 PENG Qing-yan,ZHAO Xun-jie,CHEN Jia-bo. Adaptive Window Object Tracking for Particle Filter[J].Infrared Technology,2012,34(7):568.
[2]杨悦,刘兴淼,郭启旺,等.基于改进互信息的红外目标匹配跟踪算法[J].红外技术,2013,35(06):350.
 YANG Yue,LIU Xing-miao,GUO Qi-wang,et al.Infrared Object Matching Tracking Algorithm Based on Improved Mutual Information[J].Infrared Technology,2013,35(7):350.
[3]唐耀飞,李杰.基于模板相关匹配的红外目标跟踪FPGA算法实现[J].红外技术,2012,34(03):173.
 TANG Yao-fei,LI Jie.The Implementation of FPGA Based?on Template-matching Infrared Target Tracking Algorithm[J].Infrared Technology,2012,34(7):173.
[4]王明阳,裴浩东.基于TMS320F2812的多传感器目标跟踪的控制系统设计[J].红外技术,2011,33(12):687.
 WANG Ming-yang,PEI Hao-dong.Design of the Control and Fusion System of Target Tracking withMultiple Distributed Sensors Based on TMS320F2812[J].Infrared Technology,2011,33(7):687.
[5]陈义,孙小炜,李言俊.基于人眼视觉非均匀特性的实时粒子滤波跟踪方法[J].红外技术,2010,32(11):621.
 CHEN Yi,SUN Xiao-wei,LI Yan-jun.Real-time Particle Filter Tracking Method Based on Human VisionNon-uniform Characteristics[J].Infrared Technology,2010,32(7):621.
[6]张建伟,唐 黎.扩展目标跟踪中的自适应模板刷新算法研究[J].红外技术,2014,36(4):315.[doi:10.11846/j.issn.1001_8891.201404012]
 ZHANG Jian-wei,TANG Li.Adaptive Template Update Algorithm Research for Extended Object Tracking[J].Infrared Technology,2014,36(7):315.[doi:10.11846/j.issn.1001_8891.201404012]
[7]毛丽民,卢振利,浦宇欢,等.基于FPGA实现粒子滤波算法的实时细胞跟踪系统设计[J].红外技术,2014,36(5):389.[doi:10.11846/j.issn.1001_8891.201405010]
 MAO Li-min,LU Zhen-li,PU Yu-huan,et al.Design of a Real-time Cell Tracking System by Implementation of Particle Filter Algorithm with FPGA[J].Infrared Technology,2014,36(7):389.[doi:10.11846/j.issn.1001_8891.201405010]
[8]管学伟,刘先志,罗镇宝.基于区域协方差矩阵的目标跟踪方法[J].红外技术,2009,31(2):099.
 GUAN Xue-wei,LIU Xian-zhi,LUO Zhen-bao.Object Tracking Algorithm Based on Region Covariance Matrix[J].Infrared Technology,2009,31(7):099.
[9]金晓会,杨 卫,白晓方.一种用红外热像仪检测与跟踪遮挡目标的方法[J].红外技术,2014,36(8):639.[doi:10.11846/j.issn.1001_8891.201408007]
 JIN Xiao-Hui,YANG Wei,BAI Xiao-Fang.A Method Using Infrared Thermal Imager to Detect and Track Hidden[J].Infrared Technology,2014,36(7):639.[doi:10.11846/j.issn.1001_8891.201408007]
[10]李树军.基于数据关联快速算法的目标跟踪与仿真研究[J].红外技术,2008,30(五):268.
 LI Shu-jun.Research on Targets Tracking Using Fast Algorithm?for Data Association and Simulation[J].Infrared Technology,2008,30(7):268.

备注/Memo

备注/Memo:
收稿日期:2019-08-27;修订日期:2020-07-09.
作者简介:沈旭(1979-),男,硕士,讲师,主要研究方向图像处理、智能控制应用、模式识别等。
基金项目:国家自然科学基金(61662017,61402399);湛江市科技发展专项资金竞争性分配项目(2019A01042);岭南师范学院教育教学改革项目(LSJGMS1811)。

更新日期/Last Update: 2020-07-16