A Bayesian tree-search track initiation algorithm for dim targets

A novel algorithm for multi-target track initiation in dense clutter environments is proposed based on approximating local maxima in the observation likelihood function. The algorithm implements a tree structure to search for local maxima of the observation likelihood function by dividing the entire surveillance area into large subsets and narrowing the search inside each subset in which there is a high likelihood that a target is present. A rough Gaussian approximation technique is proposed to reduce complexity in calculating the observation likelihood function over a subset by avoiding integration. The proposed algorithm has been tested on a multi-target benchmark dataset and shows superior performance in terms of high target detection probability, low probability of false alarm, and low computational complexity.