Abstract
In intelligent robots, three-dimensional (3-D) instrumentation and object recognition are required to move autonomously. Moreover, not only throughput but also the latency between the image acquisition and the output generation of the 3-D instrumentation and object recognition result should be considered because of its frequentt visual feedback. However, enormously large computation power is required to perform 3-D instrumentation and object recognition because they handle two-dimensional (2-D) and 3-D image. Development of a special-purpose VLSI processor for 3-D instrumentation and abject recognition is essential for real-time motion of the intelligent robots.
In this paper, a Model-Based Robot Vision (MBRV) VLSI processor for 3-D instrumentation and object recognition is proposed. The principle of its algorithm is model matching between an input image and 3-D models. Because the algorithm always gives the candidates for the accurate 3-D instrumentation and object recognition result with simple and regular procedures, it is suitable for the implementation of the VLSI processor.
Highly parallel architecture is employed in the VLSI processor to reduce the latency. Pipeline architecture is very useful to reduce the latency because a large amount of data is streamed into the VLSI processor successively. Spatial parallel architecture based on data flow graph (DFG) is also introduced for regular pipeline data flow.
As a result, it is confirmed that the proposed VLSI processor can perform 3-D instrumentation and object recognition 10000 times faster than 28.5 MIPS workstation.
