《计算机应用研究》|Application Research of Computers

多线索融合的无视野重叠跨摄像机行人跟踪算法

Pedestrian tracking across non-overlapping camera views based on fusion multi-clue

免费全文下载 (已被下载 次)  
获取PDF全文
作者 谭飞刚,黄玲,刘建
机构 1.深圳信息职业技术学院,广东 深圳 518172;2.华南理工大学,广州 510641;3.中国人民解放军95339部队,南宁 532500
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2017)08-2542-04
DOI 10.3969/j.issn.1001-3695.2017.08.068
摘要 针对单个摄像机视野有限而无法满足日益扩大的监控范围的现象,对无视野重叠的跨摄像机行人跟踪算法进行了研究,并提出了一种融合时空线索和外观线索的无视野重叠跨摄像机行人跟踪算法。在对已有摄像机网络拓扑结构估计算法分析的基础上提出了一种基于加权时间窗口的无视野重叠摄像机网络拓扑结构估计算法;然后利用朴素贝叶斯完成两种线索融合,实现不同摄像机间行人匹配和跟踪信息的传递,最终实现无视野重叠区域的跨摄像机行人跟踪。该算法在公开的MCT数据集上进行对比实验并取得了优于其他算法的结果。
关键词 行人跟踪;跨摄像机跟踪;无视野重叠;多线索融合;拓扑结构估计
基金项目 国家自然科学基金资助项目(51408237)
本文URL http://www.arocmag.com/article/01-2017-08-068.html
英文标题 Pedestrian tracking across non-overlapping camera views based on fusion multi-clue
作者英文名 Tan Feigang, Huang Ling, Liu Jian
机构英文名 1.ShenzhenInstituteofInformationTechnology,ShenzhenGuangdong518172,China;2.SouthChinaUniversityofTechnology,Guangzhou510641,China;3.95339TroopsofPLA,Nanning532500,China
英文摘要 For the phenomenon that the single camera view was limited and could not meet the growing scope of monitor, this paper studied algorithms that pedestrian tracking across non-overlapping camera views and proposed a pedestrian tracking across non-overlapping camera views based on fusing spatial-temporal and appearance clues. On the basis of analyzing the existing camera network topology estimation algorithm, it proposed the cameras network topology estimation algorithm based on weighted time window. Then it fused two kinds of clues by using naive Bayes and achieved pedestrian match and transfer of tracking information across different cameras. Lastly, it realized pedestrian tracking across non-overlapping camera views. It comparative tested the proposed algorithm in MCT dataset, and results are better than other methods.
英文关键词 pedestrian tracking; tracking across cameras; non-overlapping camera views; fusion multi-clue; topology estimation
参考文献 查看稿件参考文献
  [1] Kokura T, Kawanishi Y, Mukunoki M, et al. Tracking pedestrians across multiple cameras via partial relaxation of spatio-temporal constraint and utilization of route cue[C] //Proc of ACCV Workshops. [S. l. ] :Springer International Publishing, 2014:587-601.
[2] Wang Lu, Yung N H, Xu Lisheng. Multiple-human tracking by iterative data association and detection update[J] . IEEE Trans on Intelligent Transportation Systems, 2014, 15(15):1886-1899.
[3] Chen K W, Lai C C, Lee P J, et al. Adaptive learning for target tracking and true linking discovering across multiple non-overlapping cameras[J] . IEEE Trans on Multimedia, 2011, 13(4):625-638.
[4] Chen Xiaotang, Huang Kaiqi, Tan Tieniu. Object tracking across non-overlapping views by learning inter-camera transfer models[J] . Pattern Recognition, 2014, 47(3):1126-1137.
[5] Niu Chaowei, Grimson E. Recovering non-overlapping network topo-logy using far-field vehicle tracking data[C] //Proc of International Conference on Pattern Recognition. 2006:944-949.
[6] Chen Xiaotang, Huang Kaiqi, Tan Tieniu. Learning the three factors of a non-overlapping multi-camera network topology[M] //Pattern Recognition. Berlin:Springer, 2012:104-112.
[7] Javed O, Rasheed Z, Shafique K, et al. Tracking across multiple cameras with disjoint views[M] . Los Alamitos:IEEE Computer Society, 2003:952-957.
[8] Makris D, Ellis T, Black J. Bridging the gaps between cameras[C] //Proc of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2004:205-210.
[9] Pham N T, Leman K, Chang R, et al. Fusing appearance and spatio-temporal features for multiple camera tracking[M] // MultiMedia Modeling. [S. l. ] :Springer International Publishing, 2014:365-374.
[10] Zhao Rui, Ouyang Wanli, Wang Xiaogang. Learning mid-level filters for person re-identification[C] //Proc of IEEE Conference on Computer Vision and Pattern Recognition. 2014:144-151.
[11] Nam Y, Rho S, Park J H. Inference topology of distributed camera networks with multiple cameras[J] . Multimedia Tools & Applications, 2013, 67(1):289-309.
[12] 谭飞刚, 刘伟铭. 多部件验证的双层行人检测算法[J] . 华南理工大学学报:自然科学版, 2015, 43(1):79-86.
[13] 谭飞刚, 刘伟铭, 黄玲, 等. 基于加权欧氏距离度量的目标再识别算法[J] . 华南理工大学学报:自然科学版, 2015, 43 (9):88-94.
[14] Han M, Kim I. Hue modeling for object tracking in multiple non-overlapping cameras[M] // Multi-disciplinary Trends in Artificial Intelligence. Berlin:Springer, 2013:69-78.
[15] Teixeira L F, Corte-Real L. Video object matching across multiple independent views using local descriptors and adaptive learning[J] . Pattern Recognition Letters, 2009, 30(2):157-167.
[16] Yang Dikai, Chung P C, Huang Chunrong. Unsupervised path mo-deling across multiple cameras with disjoint views for foreground object tracking[C] //Proc of International Conference on Information Science, Electronics and Electrical Engineering. 2014:1161-1165.
[17] Tan Feigang, Liu W M, Huang L, et al. Person re-identification across multiple non-overlapping cameras by grouping similarity comparison model[J] . Chinese Journal of Electronics, 2016.
[18] Chen Xiaojing, An Le, Bhanu B. Multitarget tracking in nonoverlapping cameras using a reference set[J] . IEEE Sensors Journal, 2015, 15(5):2692-2704.
收稿日期 2016/5/25
修回日期 2016/7/15
页码 2542-2545
中图分类号 TP391.41
文献标志码 A