OPTIMAL ENCODING OF GRAPH HOMOMORPHISM ENERGY USING FUZZY INFORMATIONAGGREGATION OPERATORS

The attributed relational graph matching (ARG) strategy is a well-know n approach to object/pattern recognition. In the past for the parallel solution of ARG matching problem, an overall objective function was c onstructed using linearly weighted information aggregation function an d one set of parameter values was chosen for all models by trial-and-e rror for the parameters in the function. In this paper, the compatibil ity between every pair of model and scene attributes is interpreted as a fuzzy value and subsequently the nonlinear fuzzy information aggreg ation operators are used to fuse the information captured in the chose n attributes. To learn the parameters in the fuzzy information aggrega tion operators, the ''learning from samples'' strategy is used. The le arning of weight parameters is formulated as an optimisation problem a nd solved using the gradient projection algorithm based learning proce dure. The learning procedure implicitly evaluates ambiguity, robustnes s and discriminatory power of the relational attributes chosen for gra ph matching and assigns weights appropriately to the chosen attributes . The learning procedure also enables us to compute a distinct set of optimal parameters for every model to reflect the characteristics of t he model so that the homomorphic ARG matching problem can be optimally encoded in the energy function for the model. Experimental results ar e presented to illustrate effectiveness and necessity of the parameter estimation and learning procedures. (C) 1998 Pattern Recognition Soci ety. Published by Elsevier Science Ltd. All rights reserved.