Classification is a basic task in data mining and pattern recognition that requires the construction of a classifi- er

that is

a function that assigns a class label to instances described by a set of features（or attributes）.Recently

a lot of new methods come forth

such as Fuzzy sets

Rough sets

Neural Network

Support Vector Machine and Genetic Algorithms

Ant Behavior Simulation

Case-based Reasoning

Bayesian Network etc..Since 1988 Pearl et al.provided the concept of Bayesian networks for the first time

it has been popular used in the Artificial Intelligence（AI）communi- ty.Bayesian networks are powerful formalism for representing and reasoning under conditions of uncertainty

and it is a powerful tool for knowledge representation and inference under conditions of uncertainty.However

Bayesian Network was not considered as classifiers until the discovery of Naive Bayesian Network

a simple kind of Bayesian Networks

which assumes the independent features given the class attribute（node）and is surprisingly effective.From that time on

it trig- gered experts to explore more deeply into Bayesian Networks as classifiers.Since the"Naive"independent assumption in Naive Bayesian Network can not be held in many cases

researchers have wondered whether the performance will be better if the strong independent assumption among features（or variables）were relaxed.In this paper

on the basis of the study of Naive Bayes Classifiers

the Na’lve Bayesian Network is generalized and is united with maximum likelihood classifier （MLC）on the mathematic model.To validate the feasibility and effectivity of the proposed method applied in the texture classification of aerial image

some aerial images of some cities in China are used in the experiments.The experiment results demonstrate that the generalized Bayesian Network-n-level Bayesian Network performs better in overall classifica- tion precision than maximum likelihood classifier and Naive Bayesian Network.The proposed method considers correlations among features

but the correlations are inherent.It is difficult to represent the correlations only by the parameter n.Dif- ferent values of n can get different experiment results

which will be researched deeply in the future.