Data Fusion Enhanced with Context Information for Road Safety Application

Abstract

Traffic accidents are an important socio-economic problem. Every year, their cost to human lives and the economic consequences are inestimable. Efforts to reduce or mitigate this problem have led to a reduction of the death toll in the past years. However, this number is still significant, thus further actions are necessary. Recent advances in computational and information technologies have led to sophisticated applications, which have the ability to warn the driver, or even to take control of the vehicle in case of danger, providing a safer and more efficient driving experience. These advanced and complex systems demand trustable and accurate sensing technologies that are able to provide environment perception and identify all the agents on it. The available sensing technologies are insufficient to provide this accurate environment perception, thus the combination of different technologies available is mandatory in order to fulfill the demanding requirements of these applications. Furthermore, within this scope, contextual information (CI) has a key role to provide complete situation assessment. This chapter describes a road safety application for pedestrian detection, using both laser scanner and optical sensor technologies. CI is used to enhance these detections and provide situation assessment. Both online and offline information are used. Online information is used to obtain the status of the vehicle and its localization. Offline information allows to understand the interaction with the detected targets, i.e., danger involving the interaction with the different users of the roads can be estimated, thanks to the information regarding traffic safety, physical behavior, road localization, etc. Furthermore, CI is also used to enhance the detections, thanks to information such as anthropometric data, pedestrian movement information, etc. The proposed method provides multilevel solution for road safety application, and is able to provide trustable detection, situation assessment, and threat detection.