Bayesian inference of thermodynamic state incorporating Schwarz-Rissanen complexity for infrared target recognition

The recognition of targets in IR scenes is complicated by the wide variety of appearances associated with different thermodynamic states. We represent variability in the thermal signatures of targets via an expansion in terms of "eigentanks" derived from a principal component analysis performed over the target's surface. Employing a Poisson sensor likelihood, or equivalent ly a likelihood based on Csiszar's I-divergence (a natural discrepancy meas ure for nonnegative images), yields a coupled set of nonlinear equations wh ich must be solved to compute maximum a posteriori estimates of the thermod ynamic expansion coefficients. We propose a weighted least-squares approxim ation to the Poisson loglikelihood for which the MAP estimates are solution s of linear equations. Bayesian model order estimation techniques are emplo yed to choose the number of expansion coefficients; this prevents target mo dels with numerous eigentanks in their representation from having an unfair advantage over simple target models. The Bayesian integral is approximated by Schwarz's application of Laplace's method of integration; this techniqu e is closely related to Rissanen's minimum description length criteria. Our implementation of these techniques on Silicon Graphics computers exploits the flexible nature of their rendering engines. The implementation is illus trated in estimating the orientation of a tank and the optimum number of re presentative eigentanks for both simulated and real data. (C) 2000 Society of Photo-Optical Instrumentation Engineers.