A new nonlinear regression model used for multisource-multisensor data fusion: an application of nonlinear integrals and genetic algorithms

摘要

The linear multiregression model fails in many real problems due to some inherent interaction among attributes. sources or sensors, that cannot be ignored. This interaction is totally different from the concept of correlation in statistics. To describe this kind of interaction, nonadditive set functions are used in data fusion recently. In this case, a nonlinear integral with respect to a nonadditive set function has to be used to replace the classical Lebesgue integral and establish a new nonlinear multiregression model. The regression coefficients involve the values of the set function on the power set of the attribute set and, when a proper data are available, they can be determined through a specially designed adaptive genetic algorithm where a double optimization technique is adopted.Such a nonlinear multiregressionmodel has a wide applicability in multisource-multisensor data fusion.