3-D object detection based-on features of local concavo-convex structure by coded multi flash imaging

This paper introduces an object detection method using features described information about a concavo-convex shape of an object that are obtained by using a small camera that controls the illumination direction. An image of the binarized gradient feature containing information about the shape of the object is generated by the Coded Multi Flash Imaging. The proposed method can reliably detect a texture-less object by using this feature image in the matching process. In addition, processing time for matching is significantly reduced by using a small number of pixels. Experiments using 200 actual images confirmed that the proposed method achieves a 98.5% recognition success rate and a 607 msec processing time.