Research On Curve Feature Based Visual Shape Servoing

In the process of visual robot operation, robot often need to observe the objects in a particular shape or perspective view in order to facilitate the subsequent operation. Such as object crawling, component inspection, assembly and other applications which need to adjust the view angle of operation monitoring.To achieve the perspective control tasks, we can select the information associated with the object shape as image features to visual servoing operation. However, a lot of objects don’t have any obvious geometric image features. For similar situations, we use the overall outline curve as the image features, use Bezier curve to fitting the 3-D space contour curve, make use of the properties of Bezier curves we choose the first and end points as well as their tangent vectors as the image features. For lack of the Bezier curve, we extended the image features with B-spline curves to represent the object contour curves, clever use of the property of the rational spline, proposed a method using spline control vertices as image features. Curve features which described by the Bezier curves and B-spline curves are applied to the depth independent image Jacobian visual servoing control respectively, an adaptive unknown curve parameter estimation algorithm has been designed and use these estimated parameters to design an controller, at the same time, we analyze the stability of the controller using Lyapunov theory. We obtained an robot joint speed controller which can move the camera in an eye-in-hand vision system, and achieve the visual shape servoing with object which position is unknown.Matlab simulation and physical 6-DOF manipulator experiments use the proposed curve image features have been done to verify the asymptotic stability of the controller. The common curve image features proposed in this paper, combined the visual servo control and spline curves, it make use of the special properties of the spline curve to achieve control and estimated the unknown curve parameters in the mean while. This method expands not only the use occasions of depth independent visual servo control but also the applications of spline curve, which is very significant.