Unsupervised Learning Based Image Registration of Wind Tunnel Pressure Sensitive Paint Image

Kang Liu; Xiongwei Sun; Hailiang Shi; Xianhua Wang; Hanhan Ye; Chen Cheng; Feng Zhu; Shichao Wu

doi:10.3788/AOS231885

Acta Optica Sinica, Volume. 44, Issue 9, 0915002(2024)

Unsupervised Learning Based Image Registration of Wind Tunnel Pressure Sensitive Paint Image

Kang Liu^1,2,3, Xiongwei Sun^1,2,3、*, Hailiang Shi^1,2,3、**, Xianhua Wang^1,2,3, Hanhan Ye^2,3, Chen Cheng^1,2,3, Feng Zhu^2,3, and Shichao Wu^2,3

Author Affiliations

¹University of Science and Technology of China, Hefei 230026, Anhui, China

²Anhui Institute of Optics and Fine Mechanics, Hefei Institutes of Physical Science, Chinese Academy of Sciences, Hefei 230031, Anhui, China

³Key Laboratory of General Optical Calibration and Characterization Technology, Chinese Academy of Sciences, Hefei 230031, Anhui, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(17)

Fig. 1. Flow of PSP experiment data processing

Download full size

Fig. 2. Structure of PIR-Net

Download full size

Fig. 3. Structure of convolutional networks

Download full size

Fig. 4. PSP images of shell part. (a) Windy image; (b) windless image; (c) superimposed display image

Download full size

Fig. 5. PSP images of sheet part. (a) Windy image; (b) windless image; (c) superimposed display image

Download full size

Fig. 6. Registration error images of shell part. (a) Error images of windy image; (b) error images of Feature Matching registration; (c) error images of MI-Spline registration; (d) error images of Voxelmorph registration; (e) error images of CycleMorph registration; (f) error images of BIRGU-Net registration; (g) error images of LRN registration; (h) error images of PIR-Net registration

Download full size

Fig. 7. Registration error images of sheet part. (a) Error images of windy image; (b) error images of Feature Matching registration; (c) error images of MI-Spline registration; (d) error images of Voxelmorph registration; (e) error images of CycleMorph registration; (f) error images of BIRGU-Net registration; (g) error images of LRN registration; (h) error images of PIR-Net registration

Download full size

Fig. 8. Distribution of key points of shell part

Download full size

Fig. 9. Distribution of key points of sheet part

Download full size

Fig. 10. Maximum error positions and average error values of shell part registration results. (a) Error images of windy image; (b) error images of Feature Matching registration; (c) error images of MI-Spline registration; (d) error images of Voxelmorph registration; (e) error images of CycleMorph registration; (f) error images of BIRGU-Net registration; (g) error images of LRN registration; (h) error images of PIR-Net registration

Download full size

Fig. 11. Maximum error positions and average error values of sheet part registration results. (a) Error images of windy image; (b) error images of Feature Matching registration; (c) error images of MI-Spline registration; (d) error images of Voxelmorph registration; (e) error images of CycleMorph registration; (f) error images of BIRGU-Net registration; (g) error images of LRN registration; (h) error images of PIR-Net registration

Download full size

Table 1. Comparison of registration accuracy of shell part
View table
Table 1. Comparison of registration accuracy of shell part
Method Original Feature Matching MI-Spline Voxelmorph CycleMorph BIRGU-Net LRN PIR-Net
RMSE 0.1338 0.1544 0.1043 0.0721 0.0646 0.0715 0.0624 0.0505
NCC 0.1722 0.1451 0.1196 0.3348 0.3483 0.2443 0.3512 0.4087

Table 2. Comparison of registration accuracy of sheet part
View table
Table 2. Comparison of registration accuracy of sheet part
Method Original Feature Matching MI-Spline Voxelmorph CycleMorph BIRGU-Net LRN PIR-Net
RMSE 0.1784 0.1574 0.1423 0.1696 0.1635 0.1681 0.1182 0.0479
NCC 0.0243 0.0237 0.0398 0.0269 0.0261 0.0255 0.0392 0.0596

Table 3. TRE comparison of shell part

View table

Table 3. TRE comparison of shell part

Method	TRE /pixel
Method	Point 1	Point 2	Point 3	Point 4	Point 5	Point 6
Original	7.07	5.09	4.12	9.21	8.06	8.54
Feature Matching	4.47	4.24	2.82	5.83	5.65	5.09
MI-Spline	3.16	2.82	4.24	3.60	6.08	3.16
Voxelmorph	1.41	1.21	1.63	8.06	6.32	7.07
CycleMorph	1.20	1.07	1.41	5.09	4.56	5.04
BIRGU-Net	1.48	2.11	1.66	9.05	7.07	8.06
LRN	1.21	2.23	2.13	8.24	3.04	7.28
PIR-Net	1.13	0.71	1.10	1.33	1.41	1.27

Table 4. TRE comparison of sheet part

View table

Table 4. TRE comparison of sheet part

Method	TRE /pixel
Method	Point 1	Point 2	Point 3	Point 4	Point 5	Point 6
Original	6.08	4.12	3.60	11.18	10.29	19.10
Feature Matching	3.16	8.06	15.26	8.61	11.04	15.03
MI-Spline	13.34	15.52	19.13	18.86	14.31	9.05
Voxelmorph	7.14	5.38	4.12	13.15	12.21	18.02
CycleMorph	6.08	5.04	4.56	10.04	8.60	16.12
BIRGU-Net	5.39	4.12	4.47	10.19	9.21	14.03
LRN	4.31	3.61	2.82	8.54	4.24	6.71
PIR-Net	2.24	1.70	1.41	2.23	2.11	2.62

Table 5. Comparison of registration time for different methods
View table
Table 5. Comparison of registration time for different methods
Method Feature Matching MI-Spline Voxelmorph CycleMorph BIRGU-Net LRN PIR-Net
Registration time for shell part /s 1.76 754 0.41 0.39 0.42 0.45 0.48
Registration time for sheet part /s 1.75 761 0.43 0.42 0.43 0.47 0.51

Table 6. Impact of the number of cascaded convolutional models on registration accuracy and time

View table

Table 6. Impact of the number of cascaded convolutional models on registration accuracy and time

Number of cascading CNN models	Shell part			Sheet part
Number of cascading CNN models	RMSE	NCC	Registration time /s	RMSE	NCC	Registration time /s
1 scale (1)	0.0741	0.3370	0.41	0.1786	0.0237	0.43
2 scale (1/2, 1)	0.0658	0.3476	0.45	0.1168	0.0386	0.48
3 scale (1/4, 1/2, 1)	0.0505	0.4087	0.48	0.0479	0.0596	0.52
4 scale (1/8, 1/4, 1/2, 1)	0.0496	0.4020	0.51	0.0483	0.0588	0.54

Tools

Get Citation

Copy Citation Text

Kang Liu, Xiongwei Sun, Hailiang Shi, Xianhua Wang, Hanhan Ye, Chen Cheng, Feng Zhu, Shichao Wu. Unsupervised Learning Based Image Registration of Wind Tunnel Pressure Sensitive Paint Image[J]. Acta Optica Sinica, 2024, 44(9): 0915002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites