GA-based Homography Transformation for Vision Rectification in Robot Drawing System

摘要

A robot drawing platform supporting five degrees of freedom (x, y, and z translation, z-rotation, and pitch) of a brush-pen motion has been under development in our laboratory. The platform is aimed at studying Chinese painting and calligraphy. Both replication of existing works and rendition of new are planned. This paper describes the addition of vision-based capabilities in the platform. They include projective rectification of the executed work by an overlooking camera, and the corrective drawing actions for iterative improved drawing. For enhanced performance, rectification here is implemented automatically via GA-based Homography transformation. A demonstration of how the vision information can be used to evaluate and improve the drawing is presented.