An integrated particle filter for a finite resolution sensor

A particle filtering procedure for automatic track formation using measurements from a finite resolution sensor is proposed. Automatic track formation is performed by recursively computing the posterior probability of target existence, i.e., a measure of our belief that the track is following a target, and the posterior distribution of the target state conditional on existence. An auxiliary particle filter is developed for this purpose under quite general conditions. This algorithm requires computation of the measurement likelihood, closed-form expressions for which are generally unavailable. A procedure for computing the likelihood for the case where the target dynamic and measurement equations are linear/normal is given. Simulations are used to demonstrate the superiority of the proposed method over an existing method based on normal approximations.