Power Equipment Infrared and Visible Images Registration Based on Cultural Wolf Pack Algorithm

Hongshan Zhao; Zeyan Zhang

doi:10.3788/AOS202040.1610003

Journals >Acta Optica Sinica >Volume 40 >Issue 16 >Page 1610003 > Article

Acta Optica Sinica
Vol. 40, Issue 16, 1610003 (2020)

Power Equipment Infrared and Visible Images Registration Based on Cultural Wolf Pack Algorithm

Hongshan Zhao and Zeyan Zhang^*

Author Affiliations

College of Electrical and Electronic Engineering, North China Electric Power University, Baoding, Hebei 071003, China

show less

DOI: 10.3788/AOS202040.1610003 Cite this Article Set citation alerts

Hongshan Zhao, Zeyan Zhang. Power Equipment Infrared and Visible Images Registration Based on Cultural Wolf Pack Algorithm[J]. Acta Optica Sinica, 2020, 40(16): 1610003 Copy Citation Text

show less

Fig. 1. Flow chart of SGNMI measurement algorithm

Download full size

Saliency gradient of infrared image. (a) Original image; (b) image after saliency detection; (c) infrared image after division; (d) image after enhancing the saliency area; (e) image of saliency gradient

Fig. 2. Saliency gradient of infrared image. (a) Original image; (b) image after saliency detection; (c) infrared image after division; (d) image after enhancing the saliency area; (e) image of saliency gradient

Download full size

Fig. 3. Saliency gradient of visible image. (a) Visible image; (b) image of saliency gradient

Download full size

Fig. 4. Schematic diagram of CWPA

Download full size

Fig. 5. Experimental sample of standard registration data set. (a) Visible image; (b) infrared image

Download full size

Fig. 6. Comparison results of different algorithms. (a) SMI; (b) GWW-NMI; (c) SGNMI

Download full size

Fig. 7. Part of the standard registration test image set. (a) Visible image; (b) infrared image

Download full size

Fig. 8. Test results of blurred images. (a) MAE; (b) RMSE

Download full size

Fig. 9. Visible image set and infrared image set. (a) Visible image; (b) infrared image

Download full size

Fig. 10. Test results of the actual data set. (a) Registration time; (b) MAE; (c) RMSE

Download full size

Fig. 11. Experimental results of actual data set. (a) σ_TRE; (b) registration time

Download full size

Registration parameter	h	v	q	r
Solution space	[-1000,1000]	[-1000,1000]	[-10,10]	[0,360]

Table 1. Solution space of registration parameters

Algorithm	Parameter
CPSO	N=100, i_ter=2000, inertia weight ω=0.7, learning factor c₁=c₂=1.5, individual speed limit [-0.5, 0.5]
WPA	N=100, ferocious wolves∶scout wolves=1∶1, i_ter=2000, T_max=10, step factor S=0.1; judging distance d=0.08, update scale factor β=3
CWPA	N=100, ferocious wolves∶scout wolves=1∶1, T_max=10, threshold parameter ε=0.5, update scale factor β=3

Table 2. Parameters of optimization algorithm

Sample	MAE			RMSE			Registration time /s
Sample	GWW-NMI	SMI	SGNMI	GWW-NMI	SMI	SGNMI	GWW-NMI	SMI	SGNMI
1	0.897	1.435	0.931	1.213	2.241	1.391	0.734	4.231	1.032
2	1.293	1.692	1.125	1.479	2.693	1.592	0.823	3.328	1.143
3	0.736	1.613	0.962	0.986	2.861	1.242	0.672	4.054	1.097
4	1.043	1.973	0.947	1.435	3.173	1.374	0.743	4.426	0.969

Table 3. Registration result of standard test image set

Mean MAE			Mean RMSE			Mean registration time /s
GWW-NMI	SMI	SGNMI	GWW-NMI	SMI	SGNMI	GWW-NMI	SMI	SGNMI
1.010	1.673	1.040	1.387	2.490	1.324	0.926	3.847	1.239

Table 4. Mean value of registration results of 50 sets of standard test image sets

Parameter	Value
Resolution /pixel×pixel	384×288
Scene temperature range /℃	0--200
Temperature accuracy /%	±2
Wavelength /μm	7--13
Focus range /m	>0.6
Frame rate /Hz	8.7

Table 5. Parameters of infrared camera

Function	Expression	Feature	Solution space
Sphere	$f_{1} = \overset{D}{\sum_{i = 1}} x_{i}^{2}$	L/U	[-10,10]²
Sumsquares	$f_{2} = \overset{D}{\sum_{i = 1}} i x_{i}^{2}$	L/U	[-10,10]¹⁰⁰
Booth	$f_{3} = (x_{1} + 2 x_{2} {- 7)}^{2} + (2 x_{1} + x_{2} {- 5)}^{2}$	L/M	[-10,10]²
Quadric	$f_{4} = \overset{D}{\sum_{i = 1}} {(\overset{i}{\sum_{k = 1}} x_{k})}^{2}$	L/M	[-30,30]¹⁰⁰
Powersum	$f_{5} = \overset{D}{\sum_{i = 1}} {[(\overset{D}{\sum_{j = 1}} x_{j}^{i}) - b_{i}]}^{2}$	H/U	[-10,10]²
Zakharov	$f_{6} = \overset{D}{\sum_{i = 1}} x_{i}^{2} + {(\overset{D}{\sum_{i = 1}} 0.5 i x_{i})}^{2} + {(\overset{D}{\sum_{i = 1}} 0.5 i x_{i})}^{4}$	H/U	[-10,10]¹⁰⁰
Griewank	$f_{7} = \frac{1}{4000} \overset{D}{\sum_{i = 1}} x_{i}^{2} - \overset{D}{\prod_{i = 1}} \cos (\frac{x_{i}}{\sqrt[]{i}}) + 1$	H/M	[-600,600]¹⁰⁰
Ackley	$\begin{matrix} f_{8} = - 20 \exp (- 0.2 \sqrt[]{\frac{1}{D} \overset{D}{\sum_{i = 1}} x_{i}^{2}}) - \exp [\frac{1}{D} \overset{D}{\sum_{i = 1}} \cos (2 π x_{i})] + 20 + e \end{matrix}$	H/M	[-32,32]¹⁰⁰

Table 6. Standard test functions

Function	Algorithm	MEAN	STD	SR /%	AEN
Sphere	WPA	6.31×10^-91	8.13×10^-87	100	111.30
	CPSO	7.8×10^-188	2.1×10^-185	100	34.70
	CWPA	2.45×10^-91	8.97×10^-88	100	102.10
Sumsquares	WPA	2.16×10^-96	8.57×10^-96	100	88.46
	CPSO	1.03×10^-8	1.05×10^-8	100	33.64
	CWPA	4.84×10^-96	2.17×10^-95	100	77.56
Booth	WPA	1.32×10^-6	1.7×10^-6	100	145.50
	CPSO	0	0	100	31.24
	CWPA	1.07×10^-9	1.25×10^-9	100	87.42
Quadric	WPA	7.60×10^-84	2.21×10^-89	100	326.52
	CPSO	6.88×10⁺²	1.46×10⁺²	0	2000.00
	CWPA	6.82×10^-90	3.12×10^-90	100	226.60
Powersum	WPA	6.63×10^-95	2.30×10^-95	100	113.30
	CPSO	1.09×10^-6	1.01×10^-6	100	1172.00
	CWPA	4.84×10^-183	1.27×10^-183	100	38.42
Zakharov	WPA	3.18×10^-2	2.98×10^-1	100	391.80
	CPSO	1.20	2.20	0	2000.00
	CWPA	4.15×10^-16	2.06×10^-16	100	237.98
Griewank	WPA	1.44×10^-89	5.56×10^-89	100	243.52
	CPSO	1.03×10⁺³	2.97×10⁺²	0	2000.00
	CWPA	6.63×10^-90	1.88×10^-89	100	221.94
Ackley	WPA	9.33×10^-1	1.19	60	991.04
	CPSO	9.37×10⁺¹	1.37	0	2000.00
	CWPA	4.62×10^-10	3.29×10^-12	100	193.51

Table 7. Performance comparison of optimization algorithms

Hongshan Zhao, Zeyan Zhang. Power Equipment Infrared and Visible Images Registration Based on Cultural Wolf Pack Algorithm[J]. Acta Optica Sinica, 2020, 40(16): 1610003

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information