Remote sensing is widely used in mapping land use /land cover types and monitoring land use /land cover changes from regional to global scale.Supervised classification method is a powerful tool in extracting land cover and land use information from remotely sensed images.Although many supervised classification method have been developed in machine learning field

there are not a universal best performing method yet.That is

different kinds of classification methods have their own advantages and defects.This phenomenon is called selective superiority.It is necessary to explore a method that can integrate advantages of different classifiers and avoid their weakness.Combining classifiers properly may improve classification accuracy

because different classifiers may have different mistake sets.Combined classifiers have been studied widely in machine learning field;however

it was seldom studied in remote sensing image classification.This paper proposed one type of combined classifier based on error analysis

which incorporates the rule outputs of maximum likelihood classification（MLC） and support vector machine（SVM）

to achieve higher classification accuracy.MLC is the most widely used classification method in computer processing of remotely sensed images.It is based on classical statistical theory and has solid probability meanings.However

the classified accuracy of this method would be affected seriously if the training sample distribution does not follow normal distribution.SVM is a newly developed classifier

which is based on statistical learning theory.SVM is robust for small sample

and it has shown a good performance in many studies.However

the original SVM classifier is a binary classifier

which needs to be extended to a multi-class classifier through extra works.How to effectively extend binary SVM to multi-class classification is still an on-going research issue and it probably affect the performance of SVM.The new method proposed in this paper first estimates the errors of two classifiers

which are denoted by the confidence intervals of rule outputs

then combines their rule outputs with weights depending on the confidence intervals

and finally acquires a more accurate rule output.Classification experiments were conducted on case study area（Summer Palace area in Beijing）.Classification accuracies of the combined classifier and two single classifiers were compared with different sample distribution and different sample amount.And the results demonstrated that the new combined classifier can acquire a higher accuracy than other two classifiers.The results also revealed that combined classifier performs better when two classifiers are more independent.Another compared experiment was done between new combined classifier and previous combined classifier by averaging

and result also showed that new method had better performance.However

there are still some defects in the new method.Firstly

error analysis is not completely finished for the two classifiers;secondly

error analysis based on classical statistical theory would be too optimistic for MLC.Although there are some disadvantages in the new combined classifier based on error analysis

it still has shown promising potential in remotely sensed image classification.