Evidence based analysis of internal conflicts caused by disparity inaccuracy

For future driver assistance and safety systems highly reliable and comprehensive maps of the car's environment are needed. The map quality depends highly on the accuracy and the interpretation of a given set of measurements. This contribution introduces two metrics derived from Dempster-Shafer theory. By means of these metrics two different inverse sensor models are compared. This comparison is done for two given sets of data from a stereo video sensor with pixel and subpixel accuracy. The introduced metrics reflect the abilities of the inverse sensor models to build reliable and comprehensive maps. Furthermore they show the connection between these abilities, the inverse sensor model parameter and the measurement accuracy.