Robust and Efficient Analysis Method for Camera Pose, Focal Length, and Radial Distortion Estimation under a Minimal Configuration

Yi Zhang; Baoqiong Wang; Yueqiang Zhang; Biao Hu; Wenjun Chen; Xiaolin Liu; Qifeng Yu

doi:10.3788/LOP231916

Laser & Optoelectronics Progress, Volume. 61, Issue 16, 1611020(2024)

Robust and Efficient Analysis Method for Camera Pose, Focal Length, and Radial Distortion Estimation under a Minimal Configuration

Yi Zhang^1,2, Baoqiong Wang^1,2, Yueqiang Zhang^1,2、*, Biao Hu^1,2, Wenjun Chen^1,2, Xiaolin Liu^1,2, and Qifeng Yu^1,2

Author Affiliations

¹Institute of Intelligent Optical Measurement and Detection, Shenzhen University, Shenzhen 518000, Guangdong, China

²College of Physics and Optoelectronic Engineering, Shenzhen University, Shenzhen 518000, Guangdong, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Errors and efficiencies of the algorithms varying with the noise level under the ordinary configuration. (a) Mean rotation error; (b) mean translation error; (c) mean focal length error; (d) mean runtime; (e) success rate of domain optimal solution

Download full size

Fig. 2. Errors and efficiencies of the algorithms varying with the noise level under the planar configuration. (a) Mean rotation error; (b) mean translation error; (c) mean focal length error; (d) mean runtime (e); success rate of domain optimal solution

Download full size

Fig. 3. Samples of the images used in real-data experiment. (a) ETH3D dataset (Delivery Area); (b) ETH3D dataset (Electro); (c) image for the planar experiment (left camera); (d) image for the planar experiment (right camera); (e) COLMAP dataset (Gerrard Hall); (f) COLMAP dataset (Person Hall)

Download full size

Fig. 4. Ground truths and estimations of the trajectories of the cameras in COLMAP dataset. (a) Gerrard Hall; (b) person Hall

Download full size

Table 1. Results of the ordinary and general solvers for ETH3D dataset

View table

Table 1. Results of the ordinary and general solvers for ETH3D dataset

Algorithm	Delivery Area		Electro
	Mean	Median	Mean	Median
	Rotation error /（º）
Bujnak-ord	10.06	9.84	17.02	16.04
Larsson-ord	10.03	9.81	16.97	16.00
Larsson-gen	8.86	8.72	15.15	14.18
Ours-ord	8.80	8.70	14.04	13.35
Algorithm	Translation error /%
Bujnak-ord	36.89	31.32	37.86	35.66
Larsson-ord	36.92	31.27	37.85	35.62
Larsson-gen	32.83	28.02	33.26	31.82
Ours-ord	33.76	28.41	32.34	30.75
Algorithm	Focal length error /%
Bujnak-ord	15.59	14.66	32.70	26.33
Larsson-ord	15.61	14.66	33.32	26.31
Larsson-gen	13.42	13.39	22.72	21.57
Ours-ord	15.11	14.41	30.07	24.88

Table 2. Results of the planar and general solvers for real planar data

View table

Table 2. Results of the planar and general solvers for real planar data

Algorithm	Left camera		Right camera
	Mean	Median	Mean	Median
	Rotation error /（º）
Oskarsson-pla	3.53	3.16	3.31	3.07
Bujnak-pla	7.00	6.60	7.23	6.71
Larsson-gen	8.28	7.77	8.18	7.39
Ours-pla	3.53	3.06	3.24	2.98
Algorithm	Translation error /%
Oskarsson-pla	11.08	9.43	13.49	11.42
Bujnak-pla	13.65	13.38	16.25	14.29
Larsson-gen	14.13	13.35	14.35	13.48
Ours-pla	10.30	9.35	12.38	10.53
Algorithm	Focal length error /%
Oskarsson-pla	17.84	13.11	16.88	13.74
Bujnak-pla	19.43	15.40	18.94	15.86
Larsson-gen	13.94	12.82	14.57	13.08
Ours-pla	16.97	12.24	15.63	11.77

Table 3. Results of the ordinary and general solvers combined with MSAC framework for ETH3D dataset

View table

Table 3. Results of the ordinary and general solvers combined with MSAC framework for ETH3D dataset

Algorithm	Gerrard Hall		Person Hall
	Mean	Median	Mean	Median
	Runtime /ms
Bujnak-ord	0.41	0.26	0.60	0.45
Larsson-ord	0.37	0.22	0.54	0.40
Larsson-gen	0.42	0.24	0.64	0.49
Ours-ord	0.34	0.20	0.46	0.35
Algorithm	Rotation error /（º）
Bujnak-ord	0.07	0.06	0.10	0.06
Larsson-ord	0.07	0.06	0.09	0.06
Larsson-gen	0.07	0.06	0.08	0.06
Ours-ord	0.07	0.06	0.09	0.06
Algorithm	Translation error /%
Bujnak-ord	0.28	0.15	0.35	0.17
Larsson-ord	0.23	0.15	0.34	0.17
Larsson-gen	0.23	0.15	0.29	0.15
Ours-ord	0.23	0.15	0.29	0.15
Algorithm	Focal length error /%
Bujnak-ord	0.49	0.33	0.76	0.39
Larsson-ord	0.42	0.31	0.75	0.39
Larsson-gen	0.39	0.30	0.62	0.33
Ours-ord	0.42	0.27	0.61	0.34

Table 4. Results of the planar and general solvers combined with MSAC framework for real planar data

View table

Table 4. Results of the planar and general solvers combined with MSAC framework for real planar data

Algorithm	Left camera		Right camera
	Mean	Median	Mean	Median
	Runtime /ms
Oskarsson-pla	2.89	2.32	2.39	2.10
Bujnak-pla	3.51	3.81	2.09	2.24
Larsson-gen	6.70	5.92	4.25	3.78
Ours-pla	2.19	1.90	2.19	2.08
Algorithm	Rotation error /（º）
Oskarsson-pla	0.33	0.32	0.43	0.50
Bujnak-pla	0.40	0.35	0.42	0.36
Larsson-gen	0.78	0.37	0.48	0.42
Ours-pla	0.32	0.32	0.39	0.36
Algorithm	Translation error /%
Oskarsson-pla	2.03	1.40	2.54	2.26
Bujnak-pla	2.78	1.93	2.31	2.09
Larsson-gen	6.55	2.62	1.99	0.72
Ours-pla	2.03	1.40	2.11	1.58
Algorithm	Focal length error /%
Oskarsson-pla	2.23	1.97	2.94	2.70
Bujnak-pla	2.91	2.44	2.83	2.36
Larsson-gen	6.99	2.87	2.28	1.26
Ours-pla	2.21	1.97	2.67	2.18

Tools

Get Citation

Copy Citation Text

Yi Zhang, Baoqiong Wang, Yueqiang Zhang, Biao Hu, Wenjun Chen, Xiaolin Liu, Qifeng Yu. Robust and Efficient Analysis Method for Camera Pose, Focal Length, and Radial Distortion Estimation under a Minimal Configuration[J]. Laser & Optoelectronics Progress, 2024, 61(16): 1611020

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Imaging Systems

Received: Aug. 14, 2023

Accepted: Sep. 18, 2023

Published Online: Jun. 29, 2024

The Author Email: Zhang Yueqiang (yueqiang.zhang@szu.edu.cn)

DOI:10.3788/LOP231916

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology