Michael Hoetter

Evaluation method of grid based representation from sensor data

A prerequisite for driver assistance applications is a high reliable and comprehensive description of the surrounding environment. The quality of the environmental description depends on the accuracy, reliability and the interpretation of the sensors measurements. In this paper an evaluation method is introduced based on two quality metrics derived from the Dempster-Shafer Theory. Applying these quality metrics different inverse sensor models are evaluated and compared to determine which inverse sensor model represents the most accurate and concise interpretation and which evaluation method (for the global, near, middle or far field) is optimal.

Evidence based analysis of internal conflicts from inverse sensor models

For future driver assistance and safety systems high reliable and comprehensive maps of the cars environment are needed. In this contribution we introduce two new metrics derived from Dempster-Shafer theory based on ignorance and conflict. These new metrics reflect the abilities of the inverse sensor models to build reliable and comprehensive maps. By means of these metrics we compare and optimise three different inverse sensor models for a given set of data.