Quality Measures for Lidar Signal Procoessing

This paper presents a consistent derivation of quality measures for lidar signal processing used to obtain a representation of dynamic traffic scenes for driver-safety applications. These quality measures evaluate both the individual tracks and the description of the complete scene. This evaluation is crucial since subsequent processing steps, such as warnings or actions, rely on the results of the preceding signal processing. In our approach each processing step is formulated in mathematical terms and then is evaluated using a probabilistic approach. In a subsequent step the individual quality measures are incorporated into one quality measure evaluating the overall signal processing for the individual tracks. Finally, the description of the complete scene is evaluated on the basis of the individual tracks´ quality measures. This approach has been implemented and tested on real lidar data and has proved to give a reliable measure of the signal processing quality