Calculation of Error Rates for Detection of Critical Situations in Road Traffic

Abstract

This article is a contribution to the development of methods for road safety analysis. A calculation scheme is derived for error rates of critical situations detected automatically using a road side stationary detector. A situation is classified as critical, if the time to collision is below some threshold. Calculated error rates are provided on different experiments with camera based vehicle detectors. The experiements demonstrate the best case of measurement accuracy that can be achieved using state of the art automated video surveillance technology. In the experiments, the false positive rate is five and four times higher than the true positive rate. This finding leads to the conclusion, that studies known from literature, stating there is correlation between the number of near crashes and real crashes should be faced with skepticism as long as no reliable information on error rates is provided.