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

Abstract

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 equivalently a likelihood based on Csiszar's I-divergence (a natural discrepancy measure for nonnegative images), yields a coupled set of nonlinear equations which must be solved to compute maximum a posteriori estimates of the thermodynamic expansion coefficients. We propose a weighted least-squares approximation to the Poisson loglikelihood for which the MAP estimates are solutions of linear equations. Bayesian model order estimation techniques are employed to choose the number of expansion coefficients; this prevents target models 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 technique 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 illustrated in estimating the orientation of a tank and the optimum number of representative eigentanks for both simulated and real data.

Toward practical pattern-theoretic ATR algorithms for infrared imagery

“…7 Each target in a configuration is rendered with a set of increasing, yet disjoint, region numbers so that each intensity region is colored by a different number. One can easily compute the N i by counting the pixels of a common region number and the D i by summing the corresponding data pixels.…”

“…[5][6][7] By simulating a large number of radiance measurements, taken while varying environmental and internal heating parameters over reasonable ranges, we generate a population of radiance profiles to which we apply principal component analysis. For simulating radiances, we employ the PRISM software originally developed by the Keweenaw Research Center at Michigan Technological University.…”

ATR applications of minimax entropy models of texture and shape

“…In statistical decision theory, two hypotheses are characterized by probability mass functions (or densities) PA and PB . A useful characterization of the discrepancy between two distributions, and hence the ease or difficulty of choosing between them given sample data, is the relative entropy or Kullback-Leibler distance: D(PAIIPB) = PA(X)lfl (3) Q uantities of the form, as well as a related discrepancy measure called the Chernoff information, are of vital importance in the order parameter framework. In its simplest form, the framework is an application of the theory of types, which essentially considers the behavior of empirical histograms via simple counting and bounding arguments.…”