Moving object tracking on panoramic images

Abstract

Analogic CNN algorithms are combined with the special properties of a panoramic imaging device called PAL optics. The resulting system is capable to identify and track moving objects in 360 degrees field of view real-time. Image processing engine was the ALADDIN system with a 64/spl times/64 array size CNNUM chip. As the image sensor, we tested the optical input of the CNNUM chip in one experiment, and in other cases, we used a regular camera.