Neural Network Ensemble Based On Rough Sets Reduction And Its Application To Remote Sensing Image Classification

With the development and application of the remote sensing satellite, remote sensing imageclassification has attracted much attention, and computer pattern recognition has become animportant method of classification. Pattern recognition of neural network in classification ofremote sensing image has been widely used, but based on neural network can’t simplifiedinformation space dimension, could not determine the redundant information. Rough set theoryis uncertain and fuzzy knowledge tools, in a large number of data in rough set theory can extractknowledge correlation, find hidden knowledge and reveal the latent rule, through reduction toremove redundant information. In order to reduce space dimension, rough set and neural network,constructing classification model combined. This paper, by using their respective advantages,rough set and neural network combined with, to construct a classification model.Use of their respective advantages, rough set and neural network combined with, to constructa classification model. Rough reduction intensive Jane’s many methods, this paper based on thegenetic algorithm is used in the attribute reduction algorithm, this paper introduces its theoreticalbasis. Secondly, based on neural network is introduced, the knowledge system based on attributereduction of data, with BP neural network and the classification of remote sensing image throughexperiment prove the validity of the proposed method. Later, put forward a kind of intensiveJane based on rough neural network integration method, using Boosting or Bagging correlationalgorithm produces smaller individual network to individual network integration, to experimentand achieved good effect.