Extraction edges in 3D range images to subpixel accuracy

A mixture of image processing and analytic technique is used to detect facets, edges, and corners in 3D laser rangefinder images. The methods do not require prior knowledge of object models, but assume the existence of smooth and opaque surface in the scene which may be approximated by first- or second-order analytic functions. Adequate spatial resolution is also assumed. The edges and corners in the scene can be located to subpixel accuracy. Considerable amounts of computing are required and the method cannot be considered practical without special hardware. Brief descriptions of the processing steps and experimental results are given