A Pipelined Reconfigurable Architecture for Real-time Image Processing of Robot Vision Servoing

Abstract

This paper is concerned with real-time image processing for robot's feedback control in flexible manufacturing system. Vision provides the most accurate and flexible form of feedback but has the disadvantage that it requires the enormous amounts of data throughput. Field programmable gate arrays (FPGAs) offer highly flexible designs and the parallel processing power of FPGAs allows data to be processed more quickly than a general microprocessor. A implementation in pipeline architecture of specialized hardware processor, including the key question: pipeline assignment -assembled on the basis of FPGA programmable structures -is suggested to solve the problem of robotic real-time feedback control. Coupled with the ability to adapt in real time to the needs of a particular algorithm, FPGAs are a powerful and flexible tool for image processing, with significant advantages over conventional sequential computing devices.