High-Speed Correlation Filter Tracking Algorithm Based on High-Confidence Updating Strategy

Bin Lin; Ying Li

doi:10.3788/AOS201939.0415003

Acta Optica Sinica, Volume. 39, Issue 4, 0415003(2019)

High-Speed Correlation Filter Tracking Algorithm Based on High-Confidence Updating Strategy

Bin Lin^1,2 and Ying Li^1、*

¹ Shaanxi Provincial Key Laboratory of Speech and Image Information Processing, School of Computer Science, Northwestern Polytechnical University, Xi'an, Shaanxi 710129, China

² School of Science, Guilin University of Technology, Guilin, Guangxi 541004, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Tracking results and response maps corresponding to proposed algorithm on Jogging-2 sequence. (a) Tracking results; (b) response maps

Download full size

Fig. 2. Specific embodiment of high-confidence updating strategy in tracking process on Jogging-2 sequence

Download full size

Fig. 3. Flowchart of proposed algorithm

Download full size

Fig. 4. Tracking results of proposed algorithm with different parameter settings. (a) Precision plot obtained at OPE mode; (b) success plot obtained at OPE mode

Download full size

Fig. 5. Tracking results of different algorithms on 100 video sequences. (a) Precision plot obtained at OPE mode; (b) success plot obtained at OPE mode; (c) precision plot obtained at SRE mode; (d) success plot obtained at SRE mode; (e) precision plot obtained at TRE mode; (f) success plot obtained at TRE mode

Download full size

Fig. 6. Partial tracking results on eight video sequences. (a) Tiger1; (b) DragonBaby; (c) Bird2; (d) Board; (e) Panda; (f) Jogging-1; (g) Girl2; (h) Human6

Download full size

Table 1. Precision plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

View table

Table 1. Precision plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

Tracker	IV	SV	OCC	DEF	MB	FM	IPR	OPR	OV	BC	LR
Proposed	66.70	64.73	64.07	60.99	66.76	63.75	63.78	63.75	57.27	65.49	64.19
fDSST	68.39	62.82	58.46	56.72	64.82	64.23	67.24	61.60	53.29	71.13	59.28
DSST	68.01	61.72	56.89	53.20	56.85	55.04	64.45	61.15	46.29	64.54	56.62
SWCF	67.02	61.35	58.37	53.71	55.82	52.12	63.14	59.88	45.21	63.08	53.89
CN	54.28	51.07	51.44	50.16	45.70	46.30	60.38	57.14	42.84	57.06	47.01
CFLB	37.05	44.16	41.02	39.56	39.92	40.09	45.29	41.77	33.74	38.44	55.62
CSK	47.29	44.92	42.01	42.54	36.52	38.99	48.99	47.13	27.66	52.72	41.06
KCF	64.20	58.56	58.13	56.84	56.40	57.52	63.24	61.74	47.91	64.58	51.14

Table 2. Success plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

View table

Table 2. Success plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

Tracker	IV	SV	OCC	DEF	MB	FM	IPR	OPR	OV	BC	LR
Proposed	56.58	52.48	53.56	50.54	58.67	54.44	52.20	51.79	49.84	55.09	49.82
fDSST	56.78	51.08	48.36	46.77	56.30	55.49	55.00	50.17	45.76	58.58	44.61
DSST	56.11	48.59	46.10	43.43	49.20	47.10	51.00	48.28	38.48	52.40	38.94
SWCF	55.29	48.08	46.70	43.43	48.35	44.83	49.98	47.18	37.88	51.01	36.82
CN	41.55	35.94	39.63	39.61	37.85	37.76	45.49	42.10	35.08	43.88	29.45
CFLB	29.63	32.84	31.30	31.49	34.69	34.28	35.20	31.66	27.58	31.96	35.64
CSK	36.85	32.39	33.13	33.70	31.39	32.63	38.05	35.39	24.96	41.00	26.33
KCF	47.92	39.86	44.30	43.62	45.56	44.84	47.22	45.12	39.33	49.77	30.69

Table 3. Performance evaluation of proposed algorithm in different stages at OPE mode
View table
Table 3. Performance evaluation of proposed algorithm in different stages at OPE mode
Tracker Score on precision plot /% Score on success plot /% Average speed /(frame·s^-1)
Proposed (NDR & NHU) 64.01 52.16 37.6
Proposed (NHU) 66.22 54.40 88.1
Proposed 68.61 56.81 122.3

Table 4. Average tracking speed of different algorithms
View table
Table 4. Average tracking speed of different algorithms
Tracker Proposed fDSST DSST SWCF CN CFLB CSK KCF
Average speed /(frame·s^-1) 122.3 102.2 42.6 22.0 239.3 190.1 523.1 291.1

Table 5. Attributes and relevant information of eight video sequences

View table

Table 5. Attributes and relevant information of eight video sequences

Sequence	Attributes	Frame	Object size /(pixel)
Tiger1	IV, OCC, DEF, MB, FM, IPR, OPR	354	84×67
DragonBoy	SV, OCC, MB, FM, IPR, OPR, OV	113	65×56
Bird2	OCC, DEF, FM, IPR, OPR	99	73×69
Board	SV, MB, FM, OPR, OV, BC	698	173×198
Panda	SV, OCC, DEF, IPR, OPR, OV, LR	1000	23×28
Jogging-1	OCC, DEF, OPR	307	101×25
Girl2	SV, OCC, DEF, MB, OPR	1500	171×44
Human6	SV, OCC, DEF, FM, OPR, OV	792	55×18

Tools

Get Citation

Copy Citation Text

Bin Lin, Ying Li. High-Speed Correlation Filter Tracking Algorithm Based on High-Confidence Updating Strategy[J]. Acta Optica Sinica, 2019, 39(4): 0415003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Oct. 17, 2018

Accepted: Dec. 12, 2018

Published Online: May. 10, 2019

The Author Email:

DOI:10.3788/AOS201939.0415003

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Precision plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

Table 1. Precision plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

Table 2. Success plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

Table 2. Success plot values correspond to different algorithms which are used to test sets of video sequences with different attributes at OPE mode%

Table 3. Performance evaluation of proposed algorithm in different stages at OPE mode

Table 3. Performance evaluation of proposed algorithm in different stages at OPE mode

Table 4. Average tracking speed of different algorithms

Table 4. Average tracking speed of different algorithms

Table 5. Attributes and relevant information of eight video sequences

Table 5. Attributes and relevant information of eight video sequences