A Novel Robot Vision Applicable to Real-time Target Tracking

摘要

We designed a vision system with a novel architecture composed of a silicon retina, an analog CMOS VLSI intelligent sensor, and FPGA. Two basic pre-processes are done with the silicon retina: a Laplacian-Gaussian (∇2G)-like spatial filtering and a subtraction of consecutive frames. Analog outputs of the silicon retina were binarized and transferred to FPGA in which digital image processing was executed. The system was applied to real-time target tracking under indoor illumination. Namely, the center of a target object was found as the median of the binarized image. The object could be tracked within the video frame rate in indoor illumination. The system has a compact hardware and a low power consumption and therefore is suitable for robot vision.