基于可变形模型的目标跟踪算法

近年来目标跟踪技术的研究已经有了很大的进展, 但目标的遮挡和形变仍然是目标跟踪算法面临的重大挑战。针对这些问题提出了一种基于可变形模型的目标跟踪算法。首先, 利用可变形模型对跟踪目标进行表达, 该模型将目标分为若干子块, 目标的特征由局部子块特征和全局特征共同构成。将目标的特征和子块之间的空间关系结合起来, 给出了对目标的一个统一的相似度度量函数。然后, 在线训练一个结构化输出支持向量机作为分类器, 该分类器的输出是可变形模型中目标的结构化描述。利用该分类器可以在视频及图像序列中准确地检测到目标, 完成跟踪。通过实验比较, 该算法的跟踪性能优于其他主流跟踪算法, 尤其在目标发生遮挡和形变的时候仍能准确跟踪。

Abstract

In recent years, the technology of target tracking has been greatly developed, but occlusion and deformation of the target were still the major challenges in tracking algorithms. To address these problems, a tracking algorithm based on deformable parts model (DPM) was proposed. Firstly, DPM was used to represent the target object. DPM divided the target into several small parts. The feature of the target was composed of the local feature of each part and the global feature of the entire object, then DPM defined a uniform similar function based on the object feature and spatial relationship of each pair of parts. Secondly, a structured output support vector machine (structured SVM) was trained online as the classifier, the output of the structured SVM was the structured description of the object. The target in videos or image sequences could be tracked by the detection result of the classifier. Experimental results demonstrate that the proposed methods outperform other popular trackers, especially with the challenge of object′s occlusion and deformation.

参考文献

[1] Mei X, Ling H. Robust visual tracking and vehicle classification via sparse representation[J]. IEEE Transaction on Pattern Analysis and Machine Intelligence, 2011, 33(11): 2259-2272.

[2] Ross D A, Lim J, Lin R S, et al. Incremental learning for robust visual tracking[J]. International Journal of Computer Vision, 2008, 77(1-3): 125-141.

[3] 崔雄文, 吴钦章, 蒋平, 等. 子空间模型下的仿射不变目标跟踪[J]. 红外与激光工程, 2015, 44(2): 769-774.

Cui Xiongwen, Wu Qinzhang, Jiang Ping, et al. Affine-invariant target tracking based on subspace representation[J]. Infrared and Laser Engineering,2015, 44(2): 769-774. (in Chinese)

[4] Grabner H, Bischof H. Online boosting and vision[C]//2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR′06), 2006, 1: 260-267.

[5] Kalal Z, Mikolajczyk K, Matas J. Tracking-learning-detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(7): 1409-1422.

[6] 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..

[7] Henriques J F, Caseiro R, Martins P, et al. High-Speed tracking with kernelized correlation filters[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(3): 583-596.

[8] 邢运龙, 李艾华, 崔智高, 等.改进核相关滤波的运动目标跟踪算法[J]. 红外与激光工程,2016, 45(S1): S126004.

Xing Yunlong, Li Aihua, Cui Zhigao, et al. Moving target tracking algorithm based on improved Kernelized correlation filter[J]. Infrared and Laser Engineering, 2016, 45(S1): S126004. (in Chinese)

[9] Felzenszwalb P, Grishick R, Mcallester D. Object detection with discriminatively trained part based models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(9): 1627-1645.

[10] Tsochantaridis I, Joachims T, Hofmann T. Large margin methods for structured and interdependent output variables[J]. Journal of Machine Learning Research, 2005, 6: 1453-1484.

[11] Bordes A, Bottou L, Gallinari P, et al. Solving multiclass support vector machines with LaRank[C]// International Conference on Machine Learning, ACM 2007: 89-96.

[12] Bishop C M. Pattern Recognition and Machine Learning[M]. New York: Springer, 2006: 294-323.

[13] Felzenszwalb P F, Huttenlocher D P. Distance transforms of sampled functions[J]. Theory of Computing, 2004, 19(8): 415-428.

[14] Wu Y, Lim J, Yang M, et al. Object tracking benchmark[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1834-1848.

[15] Bao C, Wu Y, Ling H, et al. Real time robust L1 tracker using accelerated proximal gradient approach[C]//IEEE Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, 2012: 1830-1837.

马俊凯, 罗海波, 常铮, 惠斌, 周晓丹, 侯德飞. 基于可变形模型的目标跟踪算法[J]. 红外与激光工程, 2017, 46(9): 0928001. Ma Junkai, Luo Haibo, Chang Zheng, Hui Bin, Zhou Xiaodan, Hou Defei. Visual tracking algorithm based on deformable parts model[J]. Infrared and Laser Engineering, 2017, 46(9): 0928001.

基于可变形模型的目标跟踪算法

关于本站 Cookie 的使用提示

全站搜索

基于可变形模型的目标跟踪算法

相关论文

相关资讯

关于本站 Cookie 的使用提示

全站搜索