Fast Hash_LBP Moving Target Detection Algorithm Based on Hamming Distance Constraint in Complex Background

Liya QIU; Weilin CHEN; Fanming LI; Shijian LIU; Xiaoyu WANG; Linhan LI

doi:10.3788/gzxb20225109.0910003

Journals >Acta Photonica Sinica >Volume 51 >Issue 9 >Page 0910003 > Article

Acta Photonica Sinica
Vol. 51, Issue 9, 0910003 (2022)

Fast Hash_LBP Moving Target Detection Algorithm Based on Hamming Distance Constraint in Complex Background

Liya QIU^{1、2、3、*}, Weilin CHEN^1、2、3, Fanming LI^1、3, Shijian LIU^1、3, Xiaoyu WANG^1、2、3, and Linhan LI^1、2、3

Author Affiliations

¹Shanghai Institute of Technical Physics，Chinese Academy of Sciences，Shanghai 200083，China

²University of Chinese Academy of Sciences，Beijing 100049，China

³Key Laboratory of Infrared System Detection and Imaging Technology，Chinese Academy of Sciences，Shanghai 200083，China

show less

DOI: 10.3788/gzxb20225109.0910003 Cite this Article

Liya QIU, Weilin CHEN, Fanming LI, Shijian LIU, Xiaoyu WANG, Linhan LI. Fast Hash_LBP Moving Target Detection Algorithm Based on Hamming Distance Constraint in Complex Background[J]. Acta Photonica Sinica, 2022, 51(9): 0910003 Copy Citation Text

show less

Fig. 1. Algorithm flow

Download full size | View in the Article

Fig. 2. Refer to the background modeling process

Download full size | View in the Article

Fig. 3. Adaptive threshold of hash_ LBP operator

Download full size | View in the Article

Fig. 4. Detection effect after using Hamming distance constraint

Download full size | View in the Article

Fig. 5. Complex background analysis

Download full size | View in the Article

Fig. 6. Schematic of adaptive threshold suppression dynamic background

Download full size | View in the Article

Fig. 7. Longitudinal comparison results of the algorithm

Download full size | View in the Article

Fig. 8. Processing results of different algorithms in complex visible light scenes

Download full size | View in the Article

Fig. 9. Comparative experimental results of two algorithms in infrared scene

Download full size | View in the Article

Fig. 10. Comparison of R_e，P_r and F histograms of six algorithms in each scenario

Download full size | View in the Article

Fig. 11. Hash_LBP detects results in ViBe and GMM

Download full size | View in the Article

Algorithm	Cano	Overpass	Highway	Skating	Blizzard	Snowfall
CODE_BOOK	0.034 8	0.055 6	0.037 1	0.068 5	0.012 7	0.012 5
LBP_MRF	0.185 5	0.100 5	0.036 1	0.081 3	0.013 3	0.024 4
ViBe	0.051 7	0.038 7	0.039 0	0.048 6	0.010 9	0.011 9
KDE	0.170 8	0.134 0	0.130 9	0.033 4	0.016 7	0.024 2
GMM	0.062 3	0.039 5	0.073 1	0.049 3	0.020 7	0.056 9
Proposed	0.022 5	0.030 8	0.004 7	0.018 4	0.010 4	0.046 5

Table 1. False detection rate of each algorithm in different visible scenes

View in the Article

Algorithm	Corridor	Diningroom	Lakeside	Library	Park
CODE_BOOK	0.099 8	0.039 7	0.016 5	0.038 9	0.018 3
LBP_MRF	0.024 0	0.042 7	0.018 4	0.165 9	0.019 2
ViBe	0.024 1	0.063 3	0.021 0	0.202 2	0.017 9
KDE	0.024 5	0.065 0	0.019 0	0.171 0	0.018 0
GMM	0.025 8	0.060 3	0.021 2	0.198 3	0.015 7
Proposed	0.028 6	0.024 3	0.009 4	0.041 0	0.015 1

Table 2. False detection rate of each algorithm in different infrared scenes

View in the Article

Dataset	Algorithm	R_e	P_r	F	P_PWC	Dataset	Algorithm	R_e	P_r	F	P_PWC
Cano	CODE_BOOK	0.737 6	0.665 5	0.699 7	0.034 8	Blizzard	CODE_BOOK	0.467 5	0.760 4	0.579 1	0.012 7
	LBP_MRF	0.978 9	0.102 4	0.472 9	0.185 5		LBP_MRF	0.657 4	0.680 7	0.668 8	0.013 3
	ViBe	0.428 3	0.477 0	0.538 2	0.051 7		ViBe	0.587 6	0.826 5	0.686 9	0.010 9
	KDE	0.984 0	0.178 1	0.301 6	0.170 8		KDE	0.216 8	0.857 0	0.346 1	0.016 7
	GMM	0.146 2	0.153 0	0.149 5	0.062 3		GMM	0.042 3	0.426 1	0.076 9	0.020 7
	Proposed	0.801 0	0.791 8	0.796 4	0.022 5		Proposed	0.690 5	0.692 1	0.691 3	0.010 4
Lakeside	CODE_BOOK	0.352 4	0.689 5	0.466 4	0.016 5	Overpass	CODE_BOOK	0.783 5	0.403 4	0.532 6	0.055 6
	LBP_MRF	0.326 2	0.539 4	0.406 6	0.018 4		LBP_MRF	0.972 4	0.345 5	0.509 9	0.100 5
	ViBe	0.124 6	0.621 7	0.207 7	0.021 0		ViBe	0.589 6	0.656 3	0.621 0	0.038 7
	KDE	0.057 4	0.569 9	0.104 3	0.019 0		KDE	0.974 2	0.283 2	0.438 9	0.134 0
	GMM	0.205 2	0.560 0	0.300 3	0.021 2		GMM	0.602 7	0.641 2	0.621 4	0.039 5
	Proposed	0.993 2	0.635 0	0.718 2	0.009 4		Proposed	0.736 0	0.770 7	0.682 3	0.030 8
Snowfall	CODE_BOOK	0.688 9	0.798 3	0.739 6	0.012 5	Library	CODE_BOOK	0.936 0	0.891 1	0.903 7	0.038 9
	LBP_MRF	0.898 3	0.456 8	0.605 6	0.024 4		LBP_MRF	0.333 7	0.806 8	0.472 1	0.165 9
	ViBe	0.567 6	0.803 9	0.665 4	0.011 9		ViBe	0.118 8	0.808 4	0.207 1	0.202 2
	KDE	0.311 5	0.397 1	0.349 2	0.024 2		KDE	0.277 2	0.857 4	0.419 0	0.171 0
	GMM	0.134 8	0.067 5	0.089 9	0.056 9		GMM	0.154 0	0.256 8	0.771 7	0.198 3
	Proposed	0.910 3	0.813 3	0.859 1	0.046 5		Proposed	0.899 9	0.914 5	0.907 2	0.041 0
Highway	CODE_BOOK	0.774 9	0.657 8	0.711 5	0.037 1	Corridor	CODE_BOOK	0.910 6	0.248 6	0.390 2	0.099 8
	LBP_MRF	0.905 9	0.627 5	0.758 6	0.036 1		LBP_MRF	0.769 0	0.628 5	0.691 7	0.024 0
	ViBe	0.628 6	0.685 5	0.655 8	0.039 0		ViBe	0.458 4	0.741 9	0.566 6	0.024 1
	KDE	0.927 8	0.298 8	0.452 0	0.130 9		KDE	0.367 5	0.821 3	0.507 8	0.024 5
	GMM	0.498 7	0.403 8	0.446 3	0.073 1		GMM	0.622 3	0.626 8	0.624 6	0.025 8
	Proposed	0.939 2	0.697 1	0.658 1	0.004 7		Proposed	0.723 6	0.989 2	0.813 4	0.028 6
Park	CODE_BOOK	0.508 6	0.635 7	0.565 1	0.018 3	Skating	CODE_BOOK	0.624 0	0.641 3	0.632 5	0.068 5
	LBP_MRF	0.877 8	0.453 3	0.597 9	0.019 2		LBP_MRF	0.930 7	0.540 5	0.683 9	0.081 3
	ViBe	0.560 7	0.632 8	0.594 6	0.017 9		ViBe	0.579 0	0.816 0	0.677 4	0.048 6
	KDE	0.374 3	0.720 0	0.492 5	0.018 0		KDE	0.838 0	0.814 0	0.825 8	0.033 4
	GMM	0.567 8	0.703 1	0.628 3	0.015 7		GMM	0.626 5	0.770 4	0.691 0	0.049 3
	Proposed	0.607 7	0.603 7	0.605 7	0.015 1		Proposed	0.829 5	0.930 1	0.876 9	0.018 4
Dining room	CODE_BOOK	0.829 2	0.736 7	0.780 2	0.039 7
	LBP_MRF	0.680 3	0.784 2	0.728 6	0.042 7
	ViBe	0.288 7	0.878 9	0.434 6	0.063 3
	KDE	0.447 0	0.671 6	0.536 8	0.065 0
	GMM	0.370 7	0.810 3	0.508 6	0.060 3
	Proposed	0.921 4	0.827 0	0.894 9	0.024 3

Table 3. Values of R_e、P_r、F and P_PWC of different algorithms in each scenario

View in the Article

Scenes	Size	Time/ms
Scenes	Size	CODE_BO OK	LBP_MRF	Vibe	KDE	GMM	Proposed
Canoe	320×240	10.38	63.50	8.46	54.60	11.77	10.69
Overpass	320×240	10.31	20.70	8.41	51.35	8.88	15.47
Highway	320×240	10.41	53.33	8.10	46.43	8.79	13.61
Skating	540×360	12.31	94.28	12.18	52.43	11.37	20.11
Blizzard	720×480	14.86	193.62	16.61	52.54	15.81	31.47
Snowfall	720×480	15.63	229.10	19.36	55.63	15.33	31.54
Corridor	320×240	9.31	55.36	6.63	48.67	7.71	11.99
Diningroom	320×240	9.90	52.80	6.87	46.70	10.48	11.82
Lakeside	320×240	7.44	46.84	7.26	48.60	7.84	11.75
Library	320×240	7.02	52.50	12.29	46.84	15.78	11.77
Park	320×240	10.32	98.85	17.74	48.47	16.94	23.52

Table 4. Average processing time of different algorithms

View in the Article

Algorithms	CPU usage	Memory/MB
ViBe	22%	216
GMM	29%	224
CODE_BOOK	24%	229
Proposed	22%	221

Table 5. Comparison between CPU usage and memory

View in the Article

Dataset	Cano
Algorithm	R_e	P_r	F	P_PWC
ViBe	0.428 3	0.477 0	0.538 2	0.051 7
New_ViBe	0.488 0	0.621 7	0.546 8	0.044 2
GMM	0.146 2	0.153 0	0.149 5	0.062 3
New_GMM	0.608 0	0.758 3	0.674 9	0.034 6

Table 6. Comparison results under different algorithms

Download Citation

Save the article for my favorites

Paper Information