Collision avoidance using artificial retina sensor in ALV

The artificial retina sensor (ARS) which was developed at Osaka University in Japan, is applied to PRV II (POSTECH Road Vehicle II) for real time collision avoidance in high speed navigation. ARS consists of a linear CCD sensor and a dove prism rotating in front of the camera lens. Since ARS provides polar domain images directly from the scene, size and projection invariance in a polar coordinate system can be utilized directly. Thus one only has to apply an edge detection and a template matching method to the horizontal direction. Then optical-flow of moving objects is estimated to obtain 3D distance and time-to-impact informations from obstacles. To verify the validity of the the authors' proposed technique, real images are taken using an ARS mounted on PRV II and analyzed.