A fusion concept of video and communication data for VRU recognition

In this paper a multi sensor system for so called vulnerable road user recognition is presented, developed within an EU project for road safety improvement. The data fusion concept rests upon the sensors near infrared camera and wireless ranging devices, which are complemetary concerning their physical properties. By means of an object oriented approach using the Kalman Filter the introduced concept contains several special features. The ego motion of the sensor vehicle is simply compensated using odometric data and the object dynamics is described by means of an appropriate physical model. Furthermore the fusion system is designed in a way that it is independent on the type of the used sensors. First results of field tests show a promising enhancement of the detection reliability and real time capability compared to a single sensor system.