Theoretical Perspective on Aided Target Recognition Research

摘要

This survey closely examines the mathematical methods in use for the ill-posedproblems invariably confronted at the beginning stage of the automatic/aided target recognition problem, where features relevant to the ultimate recognition goal must be extracted from images of various kinds. These methods include maximum a posteriori estimation, regularization theoretic approaches, multiple deconvolution techniques, and the simulated-annealing method called stochastic relaxation and annealing. Also examined are some of the methodological and philosophical issues relevant to the image recognition task. A relatively detailed theoretical perspective on the overall problem area is conveyed. (Author)