| 398 | 7 | 157 |
| 下载次数 | 被引频次 | 阅读次数 |
由于运动目标变形、无规律运动等特性,现有跟踪方法在进行动态目标实时检测与分割时存在着较大困难,自适应跟踪精度较低。针对复杂背景下视频目标稳健跟踪问题,提出了一种融合深度学习和时空预测的鲁棒单目标跟踪方法。在传统基于SiamMask网络的深度学习框架内引入兴趣区域检测方法,提升动态目标的在线检测与分割精度;在跟踪系统中融入时空上下文目标跟踪算法,根据目标时空关系的在线学习,预测新的目标位置并对SiamMask模型进行算法校正,实现视频序列中的目标快速识别与跟踪,较好地改善了环境干扰、目标遮挡等复杂环境对跟踪精度的影响。验证结果表明,与传统方法相比,所提方法在精准度和鲁棒性方面有着较大提高,并且能保持较高的实时性。
Abstract：Due to the characteristics of moving object deformation and irregular motion, the existing tracking methods have great difficulties in real-time detection and segmentation of dynamic objects, and the adaptive tracking accuracy is low.In order to solve the problem of robust video target tracking in complex background, a robust single target tracking method combining deep learning and spatiotemporal prediction is proposed.Firstly, the region of interest detection method is introduced into the traditional deep learning framework based on siammask network to improve the online detection and accurate segmentation of dynamic objects; secondly, the spatiotemporal context target tracking algorithm is integrated into the tracking system.According to the online learning of the spatiotemporal relationship of the target, the new target position is predicted and the siammask model is corrected to realize the fast recognition and tracking of the target in the video sequence.The influence of complex environment such as environmental interference and target occlusion on the tracking accuracy is better improved.Experimental results show that compared with the traditional methods, this method has great improvement in accuracy and robustness, and maintains high real-time performance.
[1] 卢湖川，李佩霞，王栋.目标跟踪算法综述[J].模式识别与人工智能，2018,31(1):61-76.
[2] 何文乐.无迹卡尔曼滤波Cam-Shift视频监控跟踪算法改进[J].无线电工程，2021,51(2):97-103.
[3] RAUTARAY S,AGRAWAL A.Vision Based Hand Gesture Recognition for Human Computer Interaction:a Survey[J].Artificial Intelligence Review,2015,43(1):1-54.
[4] LI Y,AI H,YAMASHITA T,et al.Tracking in Low Frame Rate Video:A Cascade Particle Filter with Discriminative Observers of Different Life Spans[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2017,30(10):1728-1740.
[5] MIN W,FAN M,GUO X,et al.A New Approach to Track Multiple Vehicles with the Combination of Robust Detection and Two Classifiers[J].IEEE Trans.on Intelligent Transportation Systems,2017,19(1):174-186.
[6] LIU Q,YANG J,ZHANG K,et al.Adaptive Compressive Tracking via Online Vector Boosting Feature Selection[J].IEEE Trans.on Cybernetics,2015,47(12):4289-4301.
[7] KRISTAN M,MATAS J,LEONARDIS A,et al.A Novel Performance Evaluation Methodology for Single-target Trackers[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(11):2137-2155.
[8] CHEN K,TAO W B.Once for All:A Two-flow Convolutional Neural Network for Visual Tracking[J].IEEE Transactions on Circuits and Systems for Video Technology,2018,28(12):3377-3386.
[9] 侯志强，陈立琳，余旺盛，等.基于双模板Siamese网络的鲁棒视觉跟踪算法[J].电子与信息学报，2019,41(9):2247-2255.
[10] WU Y,LIM J W,YANG M H.Object Tracking Benchmark[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2015,37(9):1834-1848.
[11] 梁硕，陈金勇，吴金亮.简易交互式视频目标跟踪技术研究[J].无线电工程，2017,47(2):48-51.
[12] GUNDOGDU E,ALATAN A A.Good Features to Correlate for Visual Tracking[J].IEEE Transactions on Image Processing,2018,27(5):2526-2540.
[13] WANG N,ZHOU W G SONG Y B,et al.Real-time Correlation Tracking via Joint Model Compression and Transfer[J].IEEE Trans.on Image Processing,2020,29:6123-6135.
[14] MA H,ACTON S T,LIN Z.SITUP:Scale Invariant Tracking Using Average Peak-to-Correlation Energy[J].IEEE Trans.on Image Processing,2020,29:3546-3557.
[15] 周珂，张浩博，付冬梅，等.复杂环境下一中基于SiamMask的时空预测移动目标跟踪算法[J].工程科学学报，2020,42(3):127-135.
[16] 谢瑜，陈莹.通道裁剪下的多特征组合目标跟踪算法[J].系统工程与电子技术，2020,42(4):764-772.
[17] ZUO W M,WU X H,LIN L,et al.Learning Support Correlation Filters for Visual Tracking[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2019,41(5):1158-1171.
[18] 刘波，许廷发，李相民，等.自适应上下文感知相关滤波跟踪[J].中国光学，2019,12(2):265-273.
[19] ZHANG S L,LU W,XING W W,et al.Learning Scale-adaptive Tight Correlation Filter for Object Tracking[J].IEEE Trans.on Cybernetics,2020,50(1):270-283.
基本信息:
中图分类号:TP391.41
引用信息:
[1]孙炯宁,吕太之,张娟,等.融合深度学习与时空预测的目标跟踪方法[J].无线电工程,2021,51(09):909-914.
基金信息:
2019江苏省“青蓝工程”优秀教学团队(软件技术专业创新教学团队)~~
2021-09-05
2021-09-05