Context aided fusion procedure for road safety application

Abstract

Road safety applications require the most reliable and trustable sensors. Context information plays also a key role, adding trustability and allowing the study of the interactions and the danger inherent to them. Vehicle dynamics, dimensions... can be very useful to avoid misdetections when performing vehicle detection and tracking (fusion levels 0 and 1). Traffic safety information is mandatory for fusion levels 2 and 3 by evaluating the interactions and the danger involved in any detection. All this information is context information that was used in this application to enhance the capacity of the sensors, providing a complete and multilevel fusion application. Present application use three sensors: laser scanner, computer vision and inertial system, the information given by these sensors is completed with context information, providing reliable vehicle detection and danger evaluation. Test results are provided to check the usability of the detection algorithm.