Robot vision via curvature and color features of objects
Kyungho Lee, Hee‐Sung Kim · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2005
A robot has to recognize the environmental objects correctly to behave as an intelligent machine. A new scheme for object recognition was suggested in this paper. Most of objects can be discriminated through the color and shape properties. The object shape was formed by the surface flatness or curvature. The surface curvature or flatness was computed by the gradients of facet functions. The facet functions can be obtained based on the gray level values of the patches in an image. The color space of the image is transformed into HSI from RGB on each patch. Thus the feature vectors of an object image are composed of the curvature and HSI values of patches in the image.