Robotic Vision Assembly using FPGA

An artificial vision system based in the BOF method implemented in a FPGA device is presented to locate and recognize rigid work pieces within an experimental workspace area, previous location is not known by a manipulator robotized element in order to grasp it and perform an assembly process (peg in hole) in real time. As the processor element a Nexys 4 development electronic FPGA card is used with Hardware Description Language (HDL) platform for its programming, and the OV7670 camera from OmniVision is used as the imaege acquisition device. Data is acquired, conditioned and processed to get information about location and feature recognition characteristics for the assembly process within an experimental manufacturing cell environment containing an educative robot manipulator of 6DOF. Two methods were used for comparison results resulting 50% success assembly for the first one when only using primitive descriptors of area an centroid and 96% success when BOF method was using. An assembly process was performed integrating the method in a single FPGA device. Having performed real time image acquisition and processing actions, the results shows the possibility of using this method for concurrent and fast assembly applications.