A new confidence estimator for vehicle tracking based on a generalization of Bayes filtering

In safety-critical driver assistance systems such as automatic emergency braking that require the estimation of the vehicle's environment usually a measure of confidence or probability of existence for tracked objects is required. Its purpose is to distinguish real objects from spurious objects based on artifacts within the measurement or tracking process in order to reduce the number of erroneous deployments (false alarms). We review and assess existing approaches of obtaining such measures. We propose a new method of computing a probability of existence by relaxing the underlying assumption of a Bayes filter which leads to a novel estimation algorithm for a probability of existence. The benefits of this approach compared to a standard Bayes filter are illustrated and corroborated by a numerical study using experimental data.