Study on remote sensing image classification based on artificial immune system

On the base of analyzing the disadvantages of the negative selection algorithm, a bidirectional selection algorithm based on artificial immune system is presented. Clone selection and mutation algorithms are used on the training sample set to obtain mature antibody set. Then the mature antibody set can be used to classify the remote sensing image. It is demonstrated that this algorithm is superior to conventional maximum likelihood classification. Its accuracy reaches 87.4 percent.