We will develop a manipulator system that can handle randomly located, moving parts. Vision algorithm to detect the planar motion of a part is implemented on an FPGA-based realtime vision system. Trajectory of a manipulator is replanned in realtime according to the detected location of a part. We will present the control architecture and will show experimental results,.

In this article, we develop parallel processing of one-sided Radon transform for the detection of the position and orientation of a planar motion object. Detection of planar motion based on one-sided Radon transform can be applied to arbitrary shapes and is robust against overlapping of objects, but it requires much computation time. Thus, we construct a parallel computation of one-sided Radon transform to reduce the computation time. The developed parallel algorithm is evaluated by comparing it with the original algorithm.