An embedded system for autonomous collision avoidance and line tracking using artificial CMOS retina sensor

In intelligent vehicle research, image processing often focuses on post-treatment. Real-time image processing needs a lot of computing. Complicated algorithms are always a challenge for any powerful computation system. This problem is often the bottleneck of system speed. By means of CMOS technology, some popular image algorithms are implemented at the hardware level. The retina sensor is a CMOS circuit including the image sensor part and some hardware implementation algorithms. Our system is based on two artificial multi-mode programmable retina sensors. These retinas are a CMOS image sensors developed at IEF and INT institutes (Institut d´Electronique Fondamentale (IEF), Universite Paris XI, Orsay-France, and Institut National de Telecommunications Evry-France) and includes some silicon analog and digital image processing algorithms. Some interesting outputs from each retina are used in an original experiment platform for autonomous collision avoidance and line tracking. We describe our embedded system, some customized characterization and some experiences using this system