Early Detection of Hazards in Driving Situations through Multi-Sensor Fusion

Abstract

This article proposes a novel approach for detecting hazardous events in driving situations. The assessment of hazards is based on the detection of atypical situations. The main assumption is that driving situations might get dangerous when an implicit normal state is not given any more. A prototype for the detection of atypical driving situations has been developed. In a first step, a multi-sensor multi-level fusion framework is presented and exemplified by object detection based on a camera and a laser scanner. The detected objects with their specific behaviours are transformed into a Statistical Information Grid (SIG), which is filled up with training information. In the working phase, current situations are compared to the statistical information and declared as atypical if under a given threshold. The prototype has been implemented. It has been shown that the recognition of atypical driving events is possible in selected driving situations.