Motion estimation of planar curves and their alignment using visual servoing

Official URL: http://192.168.1.20/record=b1293851 (Table of Contents)

Motion estimation and vision based control have been steadily improving research areas recently. Visual motion estimation is the determination of underlying motion parameters by using image data. Visual servoing on the other hand refers to the closed loop control of robotic systems using vision. Solving these problems with objects that have simple geometric features, such as points and lines is rather easy. However, these problems may imply certain challenges when we deal with curved objects that lack such simple features. This thesis proposes novel vision based estimation and control techniques that use object boundary information. Object boundaries are represented by planar algebraic curves. Decomposition of algebraic curves are used to extract features for motion estimation and visual servoing. Motion estimation algorithm uses the parameters of line factors resulting from the decomposition of the curve whereas visual servoing method employs the intersections of lines. Motion estimation algorithm is verified with several simulations and experiments. Visual servoing algorithm developed for the arbitrary alignment of a planar object is tested both with simulations on a 6 DOF Puma 560 robot and experiments on a 2 DOF SCARA robot. Results are quite promising.