Model-based robot vision VLSI processor for 3-D instrumentation and object recognition

In robot vision system, enormously large computation power is required to perform three-dimensional (3-D) instrumentation and object recognition. Moreover, many kinds of complex operations are usually required to make accurate instrumentation and recognition. In this paper, a model-based robot vision (MBRV) algorithm and its VLSI processor architecture are proposed for 3-D instrumentation and object recognition. The input image is compared with the two-dimensional (2-D) silhouette images which are generated from the 3-D object models by means of perspective projection. Because the MBRV algorithm always gives the candidates for the accurate instrumentation and recognition result with simple and regular procedures, it is suitable for the implementation of the VLSI processor. Highly parallel VLSI architecture for the algorithm is also proposed in the VLSI processor to reduce the latency equivalent to the time lapse between arrival of the input image and appearance of the result for the 3-D instrumentation. As a result, the latency of the VLSI processor becomes 10000 times faster than that of a 28.5 MIPS workstation