Detection of 3D-Flow by Characteristic of Convex-concave and Color

This paper describes a method for detecting 3D-flow of an arbitrary point on a free-form surface from time series range images obtained by a stereovision system. The approach that positively uses surface shape data obtained from the range images is examined. Especially, it proposes height-color-histogram (HCH) that integrates local convex-concave shape and color texture information on the surface as feature value used for correspondence point search between frames. HCH makes the feature of an object part integrate shape and color texture information. In this paper, first, the method of generating HCH is explained, and effectiveness is considered. Finally, it is shown that 3D-flows detection about an arbitrary point on free-form surface is possible by the proposal technique with quasi-real-time