Nonparametric Maximum Likelihood Estimation of Features in Spatial Point Processes Using Voronoï Tessellation

Abstract

This article addresses the problem of estimating the support domain of a bounded point process in presence of background noise. This situation occurs, for example, in the detection of a minefield from aerial observations. A maximum likelihood estimator for a mixture of uniform point processes is derived using a natural partition of the space defined by the data themselves: the Voronoi tessellation. The methodology is tested on simulations and compared to a model-based clustering technique.