An Efficient Region of Interest Generation Technique for Far-Infrared Pedestrian Detection

In recent years, vision systems, including thermal night vision systems, have become a standard part of modern luxury automobiles. The next generation of automotive thermal imaging systems will be expected to automate the detection of pedestrians, to enhance safety. In this paper, we outline the development of a system that will extract regions of interest (ROI) from automotive thermal imagery as part of such an automated system, while placing emphasis on low processing requirements, a typical restriction of automotive embedded systems.