• Acta Optica Sinica
  • Vol. 38, Issue 6, 0612004 (2018)
Wenjie Zhu, Guanglong Wang*, Jie Tian, Zhongtao Qiao, and Fengqi Gao
Author Affiliations
  • Laboratory of Nanotechnology and Micro System, Army Engineering University, Shijiazhuang, Hebei 050003, China
  • show less
    DOI: 10.3788/AOS201838.0612004 Cite this Article Set citation alerts
    Wenjie Zhu, Guanglong Wang, Jie Tian, Zhongtao Qiao, Fengqi Gao. Detection of Moving Objects in Complex Scenes Based on Multiple Features[J]. Acta Optica Sinica, 2018, 38(6): 0612004 Copy Citation Text show less
    Overall block diagram of algorithm
    Fig. 1. Overall block diagram of algorithm
    Color feature extraction of multi-thresholds moving objects. (a) Original image; (b) PC; (c) SC1; (d) SC2; (e) SC3; (f) SC4
    Fig. 2. Color feature extraction of multi-thresholds moving objects. (a) Original image; (b) PC; (c) SC1; (d) SC2; (e) SC3; (f) SC4
    Schematic of color and texture based multi-thresholds combined with motion compensation
    Fig. 3. Schematic of color and texture based multi-thresholds combined with motion compensation
    Compensation of color and texture combined foreground objects
    Fig. 4. Compensation of color and texture combined foreground objects
    Flow chart of Canny edge detection method combined with Canny thoughts
    Fig. 5. Flow chart of Canny edge detection method combined with Canny thoughts
    Ghost suppression effect in edge detection
    Fig. 6. Ghost suppression effect in edge detection
    Moving objects detection results obtained by several algorithms under different scenes from CDNet2014 dataset
    Fig. 7. Moving objects detection results obtained by several algorithms under different scenes from CDNet2014 dataset
    P-R curves obtained by different algorithms under six scenes from CDNet2014 dataset. (a) Bus station; (b) badminton; (c) sidewalk; (d) sofa; (e) PETS2006; (f) wet snow
    Fig. 8. P-R curves obtained by different algorithms under six scenes from CDNet2014 dataset. (a) Bus station; (b) badminton; (c) sidewalk; (d) sofa; (e) PETS2006; (f) wet snow
    Fm value in each test dataset for different algorithms
    Fig. 9. Fm value in each test dataset for different algorithms
    Foreground and mistakenly detected pixels for the proposed algorithm under different scenes. (a) Bus station; (b) badminton; (c) sidewalk; (d) sofa; (e) PETS2006; (f) wet snow
    Fig. 10. Foreground and mistakenly detected pixels for the proposed algorithm under different scenes. (a) Bus station; (b) badminton; (c) sidewalk; (d) sofa; (e) PETS2006; (f) wet snow
    Experiment. (a) Experimental devices; (b) results of object detection obtained by each algorithm
    Fig. 11. Experiment. (a) Experimental devices; (b) results of object detection obtained by each algorithm
    AlgorithmfFPRfFNRfPWCRePrFm
    DPGMM0.0070.3361.9440.6640.7410.686
    LBP0.0060.4472.0150.5530.8210.625
    MRF0.0030.4752.0440.5250.8420.623
    ViBe0.0200.2983.0830.7020.6270.619
    IMBS0.0030.3411.4310.6590.8780.730
    Proposed0.0040.3071.7840.6940.8680.754
    Table 1. Overall average performance comparison for different algorithms under six scenes from CDNet2014 dataset
    AlgorithmDPGMMLBPMRFViBeIMBSProposed
    Speed /(frame·s-1)29.82.6917.323.621.728.4
    Table 2. Processing speed comparison among different algorithms
    Wenjie Zhu, Guanglong Wang, Jie Tian, Zhongtao Qiao, Fengqi Gao. Detection of Moving Objects in Complex Scenes Based on Multiple Features[J]. Acta Optica Sinica, 2018, 38(6): 0612004
    Download Citation