Switching Probability for Track Gating in Track-to-Truth Assignment

In track-to-truth assignment applications there are tracks present which are ambiguous and appear to originate from several different truth objects. These tracks can significantly affect intermediate processes that are involved in the assignment of tracks such as bias estimation, where an incorrect track-to-truth assignment would result in a incorrect bias estimate. Gating the ambiguous tracks out of the scenario and using exclusively clear unambiguous tracks within the intermediate process enables track-to-truth assignment to be more accurate afterwards. In this work a switching probability metric is proposed that reports the ambiguity of a track report by calculating the probability of the track report switching from one truth object to another. The use of this metric for gating tracks is shown to be accurate for finding tracks and reporting ambiguity. In addition it is shown to improve bias estimation consistency when applied to track-to-truth assignment.