Weighted Correlation Filter Tracking Algorithm Based on Context and Relocation

Changzhen Xiong; Yan Lu; Jiaqing Yan

doi:10.3788/AOS201939.0415004

Journals >Acta Optica Sinica >Volume 39 >Issue 4 >Page 0415004 > Article

Acta Optica Sinica
Vol. 39, Issue 4, 0415004 (2019)

Weighted Correlation Filter Tracking Algorithm Based on Context and Relocation

Changzhen Xiong^*, Yan Lu^**, and Jiaqing Yan

Author Affiliations

Beijing Key Laboratory of Urban Intelligent Control Technologies, North China University of Technology, Beijing 100144, China

show less

DOI: 10.3788/AOS201939.0415004 Cite this Article Set citation alerts

Changzhen Xiong, Yan Lu, Jiaqing Yan. Weighted Correlation Filter Tracking Algorithm Based on Context and Relocation[J]. Acta Optica Sinica, 2019, 39(4): 0415004 Copy Citation Text

show less

Fig. 1. Overall framework of algorithm

Download full size

Fig. 2. Precision plots under different search area scales. (a) OTB-100; (b) Temple-color-128

Download full size

Fig. 3. Precision and success plots of nine algorithms on OTB-100. (a) Precision curves; (b) success curves

Download full size

Fig. 4. Comparison of typical video tracking results. (a) Human3; (b) girl2; (c) soccer; (d) lemming

Download full size

CN	HOG	Average DP /%	Average OP /%
1	0.5	85.6	79.1
1	1.0	85.7	78.2
1	1.5	84.8	78.0
1	2.0	86.8	80.3
1	2.5	84.7	77.7
Note: best results in table are in the fourth row, and results of the ECO-HC algorithm are in the second row.

Table 1. Tracking results under different feature weights

Magnification times of target size	1.1	1.2	1.3
Average CLE /pixel	19.1	19.2	23.3
Average DP /%	88.5	88.3	86.9
Average OP /%	80.0	79.7	78.6
Speed /(frame·s^-1)	65.2	65.0	64.8

Table 2. Tracking results of three contextual features

ECO-HCalgorithm	Adaptive Newton iteration	Weighted fusion	Multi-scale search area	Contextual features	Object relocation	Average DP /%	Average OP /%	Speed /(frame·s^-1)
√						85.6	78.5	60.0
√	√					85.6	78.5	66.8
√	√	√				86.8	80.3	65.7
√	√	√	√			87.7	80.4	65.8
√	√	√	√	√		88.5	80.0	65.2
√	√	√	√	√	√	89.2	80.6	65.2

Table 3. Tracking results of each strategy added in turn

	Proposedalgorithm	ECO-HCalgorithm	HCFTS	CSR-DCF	Staple	LMCF	D_LMCF	SRDCF	D_SRDCF
Average CLE /pixel	17.0	22.7	16.6	24.6	31.5	39.0	21.3	38.6	21.4
Average DP /%	89.2	85.6	87.0	79.9	78.4	78.9	85.9	78.9	85.1
Average OP /%	80.6	78.5	72.1	70.2	70.9	71.9	76.1	72.9	77.3
Speed /(frame·s^-1)	65.2	60.0	6.0	6.5	26.4	65.5	6.9	3.0	0.2

Table 4. Comparison results of 9 algorithms in OTB-100

Changzhen Xiong, Yan Lu, Jiaqing Yan. Weighted Correlation Filter Tracking Algorithm Based on Context and Relocation[J]. Acta Optica Sinica, 2019, 39(4): 0415004

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information