A biologically inspired collision detection algorithm using differential optic flow imaging

Abstract

A differential CMOS image sensor to detect collision using 1D binary differential optical flow is presented. The algorithm is inspired by the insects' use of image expansion for saccadic motion and flight landing. The proposed model is simpler in implementation than the conventional Hassenstein-Reichardt elementary motion detectors used in most collision detection models. The binary output for the optic flow is generated at the pixel level, reducing the need for digital hardware and simplifying the process of collision detection. The experimental results show that the collision detection algorithm can trigger as near as 2cm from the object, opening the possibility of using these algorithms in narrow path autonomous agent navigations.