For vision-based robotic assembly, machine vision is required to perform high-speed pattern recognitions for locating and classifying industrial parts since the visual processing capability often determines the efficiency and the effectiveness of the robotic assembly.
In the dissertation, there are presented the theory of Incremental Circle Transform and related fast algorithms of describing, locating and classifying two dimensional objects without occlusions. The incremental circle transform represents effectively the shape of boundary contour of an object as a parametric vector function of incremental elements. With the aid of similarity transform and line integral, the incremental circle transform finds the orientation of the object in a fast manner independently of the position of the object and the shift of starting point for boundary coding. Further, by adopting the concept of determinant curve of the incremental circle transform, each object is classified being invariant with respect to translation, rotation, scaling of the object and the shift of starting point for the boundary coding.
Experiments are performed by using an industrial vision system developed in the lab, and it is confirmed that the proposed algorithms are fast and robust to imaging noise compared with typical pattern recognition algorithms.