Object Classification in a High-Level Sensor Data Fusion Architecture for Advanced Driver Assistance Systems

Reliable estimation of an object´s type is an important aspect of advanced driver assistance systems (ADAS) and automated driving applications. A type-specific ADAS reaction or object prediction can therefore be realized, improving the performance of the system. Object detection research usually focuses strongly on the state and existence estimation of detected objects. In this paper, an approach is presented for estimating an the class type of an object within the framework of a high-level sensor data fusion architecture. A novel classification fusion approach using the Dempster-Shafer evidence theory is presented. The performance of the algorithms are evaluated using a test vehicle with 12 sensors for surround environment perception in an overtaking scenario on a closed test track and on the highway in real traffic.