Neural Network-Based Image Moments for Robotic Visual Servoing

Zhao, Yizhang; Xie, Wenfang; Si-ning, Liu; Wang, Tingting

doi:10.1007/s10846-014-0065-2

Cited by 13 publications

(4 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhao [22] and Mebarki [23] improved the image moment features of rotation control around the X-and Y-axes of the camera, however, the analytical expression of the Jacobian matrix using this method is still coupled in ω x and ω y direction motion control, which makes the Jacobian matrix still singular in some positions. Some scholars have proposed a feature selection method combining the advantages of neural networks [24,25] , support vector regression [26] , and other intelligent algorithms to achieve complete decoupling of ω x and ω y directions, which solves the singularity problem in visual servoing control. Manikandan [27] introduced spatial weight into image moment-based visual servoing, which solved the problem of visual servoing task failure when the target image exceeded the field of view.…”

Section: Figure 1 Efficiency Comparison Between the Handling Robot An...mentioning

confidence: 99%

Image moments-based visual servoing control of bagged agricultural materials handling robot

Han

Zuo

Chi

et al. 2023

International Journal of Agricultural and Biological Engineering

View full text Add to dashboard Cite

Manual handling is less efficient and sometimes even hazardous to humans in many areas, for example, agriculture. Using robots in those areas not only avoids human contact with such dangerous agricultural materials but also improves working efficiency. The motion of a robot is controlled using a technique called visual servoing that uses feedback information extracted from a vision sensor. In this study, a visual servoing method was proposed based on learning features and image moments for 3D targets to solve the problem of image moment-based visual servoing. A Gaussian process regression model was used to map the relationship between the image moment invariants and the rotational angles around the X-and Y-axes of the camera frame (denoted as γ and β). To obtain maximal decoupled structure and minimal nonlinearities of the image Jacobian matrix, it was assumed two image moment features, which are linearly proportional to γ and β. In addition to the other four standardized image moment features, a 6-DOF image moment-based visual servoing controller for the agricultural material handling robot was designed. Using this method, the problem of visual servoing task failure due to the singularity of the Jacobian matrix was solved, and it also had a better convergence effect for the part of the target image beyond the field of view and large displacement visual servoing system. The proposed algorithm was validated by carrying out experiments tracking bagged flour in a six-degree-of-freedom robotic system. The final displacement positioning accuracy reached the millimeter level and the direction angle positioning accuracy reached the level of 0.1°. The method still has a certain convergence effect when the target image is beyond the field of view. The experimental results have been presented to show the adequate behavior of the presented approach in robot handling operations. It provides reference for the application of visual servoing technology in the field of agricultural robots and has important theoretical significance and practical value.

show abstract

Section: Figure 1 Efficiency Comparison Between the Handling Robot An...mentioning

confidence: 99%

Image moments-based visual servoing control of bagged agricultural materials handling robot

Han

Zuo

Chi

et al. 2023

International Journal of Agricultural and Biological Engineering

View full text Add to dashboard Cite

show abstract

“…These features cannot guarantee that the Jacobian matrix has full rank all the time, leading to image singularities that might cause control instability. To avoid this problem, an efficient way is to introduce image moments as features in visual servoing (Tahri and Chaumette, 2005;Wang and Cho, 2008;Copot et al, 2009;Zhao et al, 2013;Zhao et al, 2015). Image moments have important performances like robust with respect to measurement perturbations, thus being a suitable choice as visual feedback information in servoing system (Tahri and Chaumette, 2005;Copot et al, 2009;Zhao et al, 2013;Siradjuddin et al, 2014;Zhao et al, 2015).…”

Section: Introductionmentioning

confidence: 99%

An adaptive ball-head positioning visual servoing method for aircraft digital assembly

Liu

Zhu

Dong

et al. 2019

View full text Add to dashboard Cite

Purpose To gain accurate support for large aircraft structures by ball joints in aircraft digital assembly, this paper aims to propose a novel approach based on visual servoing such that the positioner’s ball-socket can automatically and adaptively approach the ball-head fixed on the aircraft structures. Design/methodology/approach Image moments of circular marker labeled on the ball-head are selected as visual features to control the three translational degrees of freedom (DOFs) of the positioner, where the composite Jacobian matrix is full rank. Kalman–Bucy filter is adopted for its online estimation, which makes the control scheme more flexible without system calibration. A combination of proportional control with sliding mode control is proposed to improve the system stability and compensate uncertainties of the system. Findings The ball-socket can accurately and smoothly reach its desired position in a finite time (50 s). Positional deviations between the spherical centers of ball-head and ball-socket in the X-Y plane can be controlled within 0.05 mm which meets the design requirement. Practical implications The proposed approach has been integrated into the pose alignment system. It has shown great potential to be widely applied in the leading support for large aircraft structures in aircraft digital assembly. Originality/value An adaptive approach for accurate support of large aircraft structures is proposed, which possesses characteristics of high precision, high efficiency and excellent stability.

show abstract

“…Obtaining a precise nonlinear model analytically is always a hard and tedious work, even more when working with industrial systems, which do not allow knowing all the infor-mation about their systems, or simply, it is not possible to get access to the system nonlinear characteristic parameters. This is the main reason that nowadays machine learning techniques, Sun et al (1996); Kim et al (2012); Zhao et al (2015), are widely used in the field of robotics. The robot hardware is progressively becoming more complex, which leads to a growth of interest in applying machine learning and statistics approaches within the robotic community.…”

Section: Multi-rate Nonlinear Holdsmentioning

confidence: 99%

Multi-Rate Visual Feedback Robot Control

Galbis¹,

Ernesto²

View full text Add to dashboard Cite

Neural Network-Based Image Moments for Robotic Visual Servoing

Cited by 13 publications

References 21 publications

Image moments-based visual servoing control of bagged agricultural materials handling robot

Image moments-based visual servoing control of bagged agricultural materials handling robot

An adaptive ball-head positioning visual servoing method for aircraft digital assembly

Multi-Rate Visual Feedback Robot Control

Contact Info

Product

Resources

About