Target Tracking Algorithm Based on Adaptive Updating of Multilayer Convolution Features

Zeng Mengyuan; Shang Zhenhong; Liu Hui; Li Jianpeng

doi:10.3788/LOP57.021008

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 2 >Page 21008 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 2, 21008 (2020)

Target Tracking Algorithm Based on Adaptive Updating of Multilayer Convolution Features

Zeng Mengyuan, Shang Zhenhong^*, Liu Hui, and Li Jianpeng

Author Affiliations

Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming, Yunnan 650500, China

show less

DOI: 10.3788/LOP57.021008 Cite this Article Set citation alerts

Zeng Mengyuan, Shang Zhenhong, Liu Hui, Li Jianpeng. Target Tracking Algorithm Based on Adaptive Updating of Multilayer Convolution Features[J]. Laser & Optoelectronics Progress, 2020, 57(2): 21008 Copy Citation Text

show less

Fig. 1. Output images of target in VGG network. (a) Input; (b) conv3-4; (c) conv4-4; (d) conv5-4

Download full size

Fig. 2. Distance precision and overlap success curves of 8 algorithms. (a) Distance precision curves; (b) overlap success curves

Download full size

Fig. 3. Comparison of tracking results of 7 algorithms for six video sequences. (a) Football; (b) freeman3; (c) lemming; (d) shaking; (e) jogging2; (f) motorrolling

Download full size

Indicator	Ours	HCF	SRDCF	SiamFC	DSST	SAMF	KCF
FM	0.846	0.851	0.781	0.756	0.690	0.730	0.698
SV	0.889	0.875	0.874	0.825	0.784	0.748	0.770
IV	0.879	0.897	0.811	0.805	0.846	0.836	0.800
OCC	0.862	0.868	0.828	0.770	0.687	0.708	0.712
DEF	0.897	0.926	0.900	0.920	0.696	0.807	0.688
MB	0.855	0.878	0.838	0.735	0.749	0.799	0.713
IPR	0.885	0.853	0.850	0.811	0.770	0.821	0.756
OPR	0.880	0.865	0.869	0.844	0.770	0.815	0.753
OV	0.959	0.935	0.791	0.706	0.699	0.699	0.713
BC	0.905	0.905	0.811	0.901	0.695	0.802	0.706
LR	0.991	0.965	0.958	0.699	0.917	0.800	0.948

Table 1. DPs of different algorithms for different challenge factors

Indicator	Ours	HCF	SRDCF	SiamFC	DSST	SAMF	KCF
FM	0.800	0.783	0.757	0.743	0.595	0.713	0.681
SV	0.833	0.779	0.799	0.791	0.605	0.668	0.618
IV	0.833	0.860	0.765	0.783	0.666	0.782	0.647
OCC	0.814	0.798	0.785	0.745	0.578	0.685	0.611
DEF	0.882	0.879	0.830	0.892	0.561	0.770	0.592
MB	0.779	0.786	0.818	0.703	0.646	0.772	0.644
IPR	0.849	0.784	0.819	0.799	0.625	0.750	0.627
OPR	0.856	0.817	0.820	0.822	0.609	0.750	0.619
OV	0.838	0.811	0.752	0.680	0.529	0.663	0.619
BC	0.861	0.873	0.723	0.866	0.567	0.750	0.603
LR	0.809	0.834	0.776	0.640	0.666	0.738	0.686

Table 2. Success rate accuracy values of different algorithms for different challenge factors

Test video	Frame	Target size	Challenge factor
Football	362	39 ×50	OCC,IPR,OPR,BC
Freeman3	474	12× 13	SV,IPR,OPR
Lemming	1336	61 × 103	IV,SV,OCC,FM,OPR,OV
Shaking	365	61 × 71	IV,SV,IPR,OPR,BC
Jogging2	307	37×114	OCC, DEF, OPR
MotorRolling	164	122 ×125	IV,SV,MB,FM,IPR,BC,LR

Table 3. Attributes of test videos

Zeng Mengyuan, Shang Zhenhong, Liu Hui, Li Jianpeng. Target Tracking Algorithm Based on Adaptive Updating of Multilayer Convolution Features[J]. Laser & Optoelectronics Progress, 2020, 57(2): 21008

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information