• Infrared and Laser Engineering
  • Vol. 49, Issue 11, 20200284 (2020)
Lei Zhang1, Shuai Zhu2, Tianyu Liu2, and Yuehuan Wang2
Author Affiliations
  • 1Beijing Institute of Surveying and Communication, Beijing 100089, China
  • 2School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan 430074, China
  • show less
    DOI: 10.3788/IRLA20200284 Cite this Article
    Lei Zhang, Shuai Zhu, Tianyu Liu, Yuehuan Wang. Tracking of dense group targets based on motion grouping[J]. Infrared and Laser Engineering, 2020, 49(11): 20200284 Copy Citation Text show less
    Flowchart of the proposed algorithm
    Fig. 1. Flowchart of the proposed algorithm
    Dividing dense target into sparse groups
    Fig. 2. Dividing dense target into sparse groups
    Non-neighbor individuals in cluster motion
    Fig. 3. Non-neighbor individuals in cluster motion
    Relationship between and 与路径长度规模关系图
    Fig. 4. Relationship between and 与路径长度规模 关系图
    Relationship between and with regularization factor引入正则化因子后与路径长度规模关系图
    Fig. 5. Relationship between and with regularization factor 引入正则化因子后 与路径长度规模 关系图
    Distribution of group target and extracted optical flow
    Fig. 6. Distribution of group target and extracted optical flow
    Divided group result according to motion patten
    Fig. 7. Divided group result according to motion patten
    Directed graph model for target group
    Fig. 8. Directed graph model for target group
    Result of marked potential target in target group
    Fig. 9. Result of marked potential target in target group
    Track of potential target in target group
    Fig. 10. Track of potential target in target group
    Grouping performance comparison
    Fig. 11. Grouping performance comparison
    Real-time performance comparison
    Fig. 12. Real-time performance comparison
    Node123415
    10.0040.0260.039−0.27−0.06
    20.0240.0120.0570.087−0.05
    30.0360.0580.0110.004−0.03
    4−0.010.0900.004−0.03−0.05
    15−0.03−0.06−0.08−0.030.011
    Table 1.

    Result of collectiveness matrix

    聚集度矩阵结果

    Node123415
    100000
    200100
    300000
    401000
    1500000
    Table 2.

    Result of adjacency matrix for directed graph

    有向图邻接矩阵结果

    DataFrame that does not meet condition 1Frame that does not meet condition 2Accuracy
    121,4,6,8,12,14,177,9,12,16,20,23,2757,61,6255,59,60,610.8460.831
    34,8,12,13,16,17,1852,59,600.846
    45,9,18,22,3131,40,45,550.861
    59,10,11,20,20,2644,49,630.877
    Table 3.

    Tracking result without interframe suppression false alarm

    未采用帧间虚警抑制跟踪结果

    DataFrame that does not meet condition 1Frame that does not meet condition 2Accuracy
    121,4,6,8,147,9,12,16,2361,6259,60,610.8920.877
    312,13,16,17,1859,600.892
    45,3131, 45,550.923
    59,10,1136,49,630.907
    Table 4.

    Tracking result with interframe suppression false alarm

    采用帧间虚警抑制跟踪结果

    Node123415
    10.0020.0390.049−0.01−0.05
    20.0360.0100.0570.022−0.05
    30.0440.0520.0100.002−0.02
    4-0.010.0200.004−0.02−0.04
    15−0.03−0.03−0.06−0.030.013
    Table 5. Dual suppression matrix 对偶抑制矩阵
    Node123415
    10.0040.0520.049−0.02−0.06
    20.0500.0120.0570.033−0.07
    30.0460.0580.0160.002−0.03
    4−0.010.0300.004−0.03−0.05
    15−0.03−0.04−0.07−0.040.019
    Table 6. Dual suppression matrix 对偶抑制矩阵