Fusion procedure for pedestrian detection based on laser scanner and computer vision

Abstract

Sensors represent a key element in Advance Driver Assistant Systems (ADAS). The lack of reliable sensors able to fulfill the requirements for road safety applications leads to the necessity of combining them to create a fusion procedure able to overcome the limitations of a single set of sensors. Recent advances in vision based pedestrian detection have lead to robust approaches able to detect pedestrians in a wide variety of situations. On the other hand laser scanners have proved to be a very powerful tool when dealing with intelligent transport systems and safety application. By combining the capabilities of both set of sensors, a robust and trustworthy application able to deal with the most exigent necessities can be created. The presented algorithm combines the pedestrian detection given by both subsystems adding confidence to single sensor detections.