Multiple Feature Fusion based on Covariance Matrix for Visual Tracking

Zefenfen Jin; Zhiqiang Hou; Wangsheng Yu; Xin Wang

doi:10.3788/AOS201737.0915005

Journals >Acta Optica Sinica >Volume 37 >Issue 9 >Page 0915005 > Article

Acta Optica Sinica
Vol. 37, Issue 9, 0915005 (2017)

Multiple Feature Fusion based on Covariance Matrix for Visual Tracking

Zefenfen Jin^*, Zhiqiang Hou, Wangsheng Yu, and Xin Wang

Author Affiliations

Information and Navigation College, Air Force Engineering University of PLA, Xi'an, Shaanxi 710077, China

show less

DOI: 10.3788/AOS201737.0915005 Cite this Article Set citation alerts

Zefenfen Jin, Zhiqiang Hou, Wangsheng Yu, Xin Wang. Multiple Feature Fusion based on Covariance Matrix for Visual Tracking[J]. Acta Optica Sinica, 2017, 37(9): 0915005 Copy Citation Text

show less

Fig. 1. Integral image

Download full size

Fig. 2. Comparison of tracking results of different features. (a) The 649th frame of Basketball sequence; (b) the 873rd frame of Liquor sequence

Download full size

Fig. 3. Population generation method of the ith frame

Download full size

Fig. 4. Qualitative comparison of eight tracking algorithms. (a) Basketball; (b) Bolt; (c) David3; (d) Football1; (e) Jumping; (f) Liquor; (g) Matrix; (h) Skiing; (i) Ironman; (j) Jogging1; (k) Lemming; (l) MotorRolling

Download full size

Fig. 5. Center position error curves. (a) Basketball; (b) Bolt; (c) David3; (d) Football1; (e) Jumping; (f) Liquor; (g) Matrix; (h) Skiing; (i) Ironman; (j) Jogging1; (k) Lemming; (l) MotorRolling

Download full size

Fig. 6. Overlap rate curves. (a) Basketball; (b) Bolt; (c) David3; (d) Football1; (e) Jumping; (f) Liquor; (g) Matrix; (h) Skiing; (i) Ironman; (j) Jogging1; (k) Lemming; (l) MotorRolling

Download full size

Fig. 7. Curves of precision and success rate. (a) Precision; (b) success rate

Download full size

Fig. 8. Comparison of tracking precision under 11 scenes

Download full size

Fig. 9. Comparison of tracking success rate under 11 scenes

Download full size

x_i	b_i	f(x)≥f(b)	Δθ_i	s(α_iβ_i)
x_i	b_i	f(x)≥f(b)	Δθ_i	α_iβ_i>0	α_iβ_i<0	α_i=0	β_i=0
0	0	False	0	0	0	0	0
0	0	True	0	0	0	0	0
0	1	False	0	0	0	0	0
0	1	True	0.05π	-1	+1	±1	0
1	0	False	0.01π	-1	+1	±1	0
1	0	True	0.025π	+1	-1	0	±1
1	1	False	0.005π	+1	-1	0	±1
1	1	True	0.025π	+1	-1	0	±1

Table 1. Generative rules of rotation angle of quantum rotation gate

Name	DLT	TLD	Struck	ASLA	DSST	SST	MIL	Proposed algorithm
Basketball	53.9	2.3	9.93	53.9	14.3	22.2	27.5	90.1
Bolt	4.29	1.4	1.71	1.4	100.0	1.1	1.1	91.6
David3	32.9	10.7	33.73	51.6	54.0	35.7	68.3	78.6
Football1	52.4	46.0	82.4	44.6	41.9	58.1	82.4	56.7
Jumping	16.6	90.1	95.9	17.3	4.8	13.4	62.3	81.3
Liquor	20.5	56.6	41.1	23.6	40.8	23.6	20.2	91.3
Matrix	2.0	1.0	12.0	2.0	21.0	34.0	12.0	36.0
Skiing	7.4	7.4	3.7	11.1	7.4	9.9	7.4	30.4
Ironman	10.8	3.0	4.8	13.3	10.8	14.5	7.2	13.9
Jogging1	22.5	96.4	21.8	22.5	22.5	22.2	22.2	92.8
Lemming	28.0	60.5	66.2	16.9	46.0	43.0	87.2	90.8
MotorRolling	7.3	15.9	16.5	11.0	6.7	7.3	7.3	30.5

Table 2. Comparison of coverage rate of tracking results%

Name	DLT	TLD	Struck	ASLA	DSST	SST	MIL	Proposed algorithm
Basketball	12.0	-	118.3	82.6	111.6	105.9	91.9	10.8
Bolt	-	-	398.8	374.7	5.0	409.9	393.5	5.6
David3	107.4	-	106.5	87.8	88.4	104.5	29.7	15.2
Football1	10.4	-	5.5	12.2	20.5	15.7	5.6	8.9
Jumping	41.9	5.9	6.7	46.1	35.2	45.7	10.0	7.6
Liquor	153.3	-	91.0	146.7	99.3	146.7	141.9	23.2
Matrix	171.1	-	194.8	65.2	59.7	54.7	55.0	33.5
Skiing	244.5	-	251.8	266.6	220.1	269.9	267.0	89.4
Ironman	211.4	-	127.6	197.5	105.7	205.4	193.4	74.3
Jogging1	113.0	6.7	62.0	104.6	112.0	114.6	96.3	12.3
Lemming	128.9	-	37.8	178.8	81.5	23.6	12.1	11.8
MotorRolling	170.7	-	145.6	201.4	289.6	377.0	161.0	101.8

Table 3. Comparison of average center position error of tracking resultspixel

Zefenfen Jin, Zhiqiang Hou, Wangsheng Yu, Xin Wang. Multiple Feature Fusion based on Covariance Matrix for Visual Tracking[J]. Acta Optica Sinica, 2017, 37(9): 0915005

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information