A Gibbs Sampling Approach to Closely Spaced  Objects Resolution via IR Focal Plane

When track multiple closely spaced point sources using infrared sensor systems, measured signal may overlap on the focal plane, so that the sensor cannot determine the location and intensity of individual object. In this paper a novel algorithms for resolving such Closely Spaced Objects (CSO) based on Gibbs sampling and Bayesian Information Criterion (BIC) is introduced. The algorithm presents a dichotomous, iterated novel approach to resolve the CSO using the Gibbs sampling to estimate the positions and intensity signals estimation of CSO and using the Bayesian Information Criterion to detect the number of the objects. Simulations are presented to show the effectiveness of the new CSO resolution method, which based on the simulated infrared image of ballistic midcourse targets viewed by spaced IR sensor.