摘要
基于可见光图像信号对无人机等低空飞行物进行检测和追踪时,面临目标成像小、低空背景干扰多、目标速度变化快等技术难点。针对目标探测模型的性能与速度难以兼顾的问题,提出了能够保持模型性能并大幅降低运算复杂度的通用深度学习模型优化框架,实现单帧图像的目标快速检测。进一步针对目标追踪问题,利用连续检测数据,在基于探测和学习的追踪技术框架上,提出了单目标航迹追踪方法,在单帧检测结果的基础上进一步提升了准确性。测试结果显示该方法能在降低约30%计算时间的情况下仍保持运算性能,从而实现实时的高机动性目标追踪。
Low-altitude small flying object detecting and tracking based on video faces several technical challenges, including small targets, complicated background, and rapid speed changing. To solve the dilemma between performance and speed, a Convolutional Neural Network(CNN) model optimization framework was proposed, which is capable of significantly reducing computation complexity while maintaining approximate performance. To further improve the stability of object tracking in continuous frames, based on detection and learning technique framework,an optimized single object tracking method was proposed for real-time performance. Experimental results shows that the proposed method is able to reduce about 30% computation cost and maintains approximate performance as original model, which achieves real-time tracking of high mobile objects.
引文
[1] DONG X,HUANG X,ZHENG Y,et al.A novel infrared small moving target detection method based on tracking interest points under complicated background [J].Infrared Physics & Technology,2014,65:36-42.
[2] SANNA A,PRALIO B,LAMBERTI F,PARAVATI G.A novel ego-motion compensation strategy for automatic target tracking in FLIR video sequences taken from UAVs [J].IEEE Transactions on Aerospace and Electronic Systems,2009,45(2):723-734.
[3] CHUA J C,FELZENSZWALB P F.Scene grammars,factor graphs,and belief propagation [J].arXiv Preprint,2016,2016:arXiv.1606.01307.
[4] KWON J,LEE K M.Adaptive visual tracking with minimum uncertainty gap estimation [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,,2017,39(1):18-31.
[5] KRIZHEVSKY A,SUTSKEVER I,HINTON G.ImageNet classification with deep convolutional neural networks [C]// Advances in Neural Information Processing Systems,Proceedings of the 25th International Conference on Neural Information Processing Systems.Lake Tahoe:Curran Associates Inc.,2012,1:1097-1105.
[6] SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition [J].arXiv Preprint,2014,2014:arXiv:1409.1556.
[7] REN S,HE K,GIRSHICK R,SUN J.Faster R-CNN:towards real-time object detection with region proposal networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149.
[8] HUANG G,LIU Z,van der MAATEN L,et al.Densely connected convolutional networks [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2017,1:2261-2269.
[9] SI Z,ZHU S.Learning AND-OR templates for object recognition and detection [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(9):2189-2205.
[10] JADERBERG M,VEDALDI A,ZISSERMAN A.Speeding up convolutional neural networks with low rank expansions [J].arXiv Preprint,2014,2014:arXiv:1405.3866.
[11] YU X,LIU T,WANG X,TAO D.On compressing deep models by low rank and sparse decomposition [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2017,1:67-76.
[12] 王震,基于深度学习的快速目标检测技术研究[D].天津:天津理工大学,2017.
[13] 程志,赵士瑄,王伟,等.基于弱监督学习和隐藏语义分析的卷积神经网络重构[J].计算机应用,2018,38(S2):7-10.
[14] SZEGEDY C,LIU W,JIA Y,et al.Going deeper with convolutions [C]// Proceedings of the 2015 International conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society 2015:1-9.