Pose Estimation of Curved Objects Based on Binocular Vision and Vectors of the Tangent Plane

Yuzhen Liu; Jiarong Zhang; Sen Lin

doi:10.3788/LOP57.041506

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 4 >Page 041506 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 4, 041506 (2020)

Pose Estimation of Curved Objects Based on Binocular Vision and Vectors of the Tangent Plane

Yuzhen Liu¹, Jiarong Zhang^1、*, and Sen Lin^1、2、3

Author Affiliations

¹School of Electronic and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China

²State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, Liaoning 110016, China

³Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang, Liaoning 110016, China

show less

DOI: 10.3788/LOP57.041506 Cite this Article Set citation alerts

Yuzhen Liu, Jiarong Zhang, Sen Lin. Pose Estimation of Curved Objects Based on Binocular Vision and Vectors of the Tangent Plane[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041506 Copy Citation Text

show less

Fig. 1. Binocular vision measurement system model

Download full size

Fig. 2. Conversion from depth map to point cloud. (a) Depth map; (b) point cloud

Download full size

Fig. 3. Schematic of rotation of 3D coordinate system. (a) Rotate around X axis; (b) rotate around Y axis; (c) rotate around Z axis

Download full size

Fig. 4. Position representation of spatial point in world coordinate system

Download full size

Fig. 5. Pose transformation of curved object

Download full size

Fig. 6. Flow chart of the algorithm

Download full size

Fig. 7. Experimental environment

Download full size

Fig. 8. Experimental object. (a) Object 1; (b) object 2; (c) object 3; (d) object 4

Download full size

Fig. 9. Selected target corner points

Download full size

Fig. 10. Comparison of measurement errors of our algorithm and three monocular visual pose algorithms. (a) X-axis translation error; (b) Y-axis translation error; (c) Z-axis translation error; (d) rotation error around the X-axis; (e) rotation error around the Y-axis; (f) rotation error around the Z-axis

Download full size

Fig. 11. Average error percentage of our algorithm and three monocular visual pose algorithms

Download full size

Fig. 12. Comparison of measurement errors of our algorithm and two binocular visual pose algorithms. (a) X-axis translation error; (b) Y-axis translation error; (c) Z-axis translation error; (d) rotation error around the X-axis; (e) rotation error around the Y-axis; (f) rotation error around the Z-axis

Download full size

Fig. 13. Average error percentage of our algorithm and two binocular visual pose algorithms

Download full size

Data category	Translation /cm			Rotation angle /(°)
Data category	t_x	t_y	t_z	ϕ	θ	ψ
Ground truth	5.0000	4.5000	10.0000	0.0000	10.0000	10.0000
Estimated value	6.6207	4.0626	11.3933	4.2442	9.9221	10.3074
Estimation error	1.6207	0.4374	1.3933	4.2442	0.0779	0.3074
Average error	1.1505			1.5432

Table 1. Estimation results of object 1 pose change

Data category	Translation /cm			Rotation angle /(°)
Data category	t_x	t_y	t_z	ϕ	θ	ψ
Ground truth	10.0000	4.5000	0.0000	5.0000	0.0000	0.0000
Estimated value	11.3664	4.8467	0.7567	4.0535	0.5914	0.7830
Estimation error	1.3664	0.3467	0.7567	0.9465	0.5914	0.7830
Average error	0.8233			0.7736

Table 2. Estimation results of object 2 pose change

Data category	Translation /cm			Rotation angle /(°)
Data category	t_x	t_y	t_z	ϕ	θ	ψ
Ground truth	5.0000	0.0000	10.0000	0.0000	20.0000	0.0000
Estimated value	6.1803	0.1977	11.7300	4.7397	19.5883	1.8791
Estimation error	1.1803	0.1977	1.7300	4.7397	0.4117	1.8791
Average error	1.0360			2.3435

Table 3. Estimation results of object 3 pose change

Data category	Translation /cm			Rotation angle /(°)
Data category	t_x	t_y	t_z	ϕ	θ	ψ
Ground truth	0.0000	0.0000	10.0000	5.0000	0.0000	0.0000
Estimated value	0.4850	0.0327	10.0533	3.3628	0.8692	2.2322
Estimation error	0.4850	0.0327	0.0533	1.6372	0.8692	2.2322
Average error	0.1903			1.5795

Table 4. Estimation results of object 4 pose change

Object	Translation /cm	Rotation angle /(°)
1	1.1505	1.5432
2	0.8233	0.7736
3	1.0360	2.3435
4	0.1903	1.5795

Table 5. Mean estimation error of each object

Algorithm	Running time /ms
Algorithm	o_tx	o_ty	o_tz	o_ϕ	o_θ	o_ψ
COPE	11.80	8.40	8.20	8.50	8.50	8.75
ICP	595.00	382.20	408.40	536.00	618.00	621.00
Increased percentage /%	98.02	97.80	97.99	98.41	98.62	98.59
Average improved efficiency /%	98.24

Table 6. Comparison of calculation efficiency between COPE and ICP algorithm

Algorithm	Running time /ms
Algorithm	o_tx	o_ty	o_tz	o_ϕ	o_θ	o_ψ
NDT	210.00	305.00	429.20	614.00	570.00	640.00
Increased percentage /%	94.38	97.25	98.09	98.62	98.51	98.63
Average improved efficiency /%	97.58

Table 7. Comparison of calculation efficiency between COPE and NDT algorithm

Yuzhen Liu, Jiarong Zhang, Sen Lin. Pose Estimation of Curved Objects Based on Binocular Vision and Vectors of the Tangent Plane[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041506

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information