Chinese Journal of Computers   Full Text
Title	Classification of Multi-Spectral Remote Sensing Image Based on Multiple-Valued Immune Network
Authors	ZHONG Yan-Fei ZHANG Liang-Pei LI Ping-Xiang
Address	(State Key Laboratory for Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079)
Year	2007
Issue	No.12(2181—2188)
Abstract &   Background	Abstract In this paper, some initial investigations are conducted to employ multiple-valued immune network(MVIN) for classification of multi-spectral remote sensing image. The proposed method trains the immune network using the samples of regions of interest and obtains the memorial immune network. Image classification task by MVIN is attempted and the preliminary results are provided. The experiment show that the method is superior to traditional algorithms, and its overall accuracy and Kappa coefficient reach 88.84% and 0.8605 respectively. keywords artificial immune system; remote sensing; image classification; pattern recognition; immune network background Artificial immune systems (AIS) have recently drawn increased attention from the Artificial Intelligence community. AIS, which are inspired by the immune systems, use the immunological properties in order to develop adaptive systems to accomplish a wide range of tasks in various areas of research including pattern recognition, intrusion detection, clustering, optimization, and intelligence control. In spite of the successful application of AIS in several fields, few applications have been reported in remote sensing. In the previous work, the authors have proposed an unsupervised artificial immune classifier published in IEEE Transactions on Geoscience and Remote Sensing (Vol.44, No.2, 2006) and a supervised classification algorithm based on AIS that will published in the International Journal of Remote Sensing, respectively. Different with the previous work, the authors propose a new algorithm based on multiple-valued immune network to perform remote sensing image classification. In contrast to the conventional classifiers, the proposed algorithm is a self-learning highly robust algorithm. Specifically, the novelty of the algorithm lies in the following aspects: (a) it is a data driven self-adaptive method as it can adjust itself to the data without any explicit specification of functional or distributional form for the underlying model; (b) it is viewed as a universal functional approximator since it can approximate any function with arbitrary accuracy; and (c) it inherits multiple-valued logic computational capability and the memory property of multiple-valued immune network and can recognize the same or similar antigen quickly at different times. The proposed algorithm has been examined with multi-spectral image, and it is demonstrated that this algorithm can achieve high classification accuracy, thus providing an effective option for multi-spectral remote sensing image classification. This work was supported in part by the 973 Program of the People′s Republic of China under grant 2006CB701302, in part by the National Natural Science Foundation of China under grant 40471088 and 40523005. The purpose of the National Natural Science Foundation of China under Grant 40471088, named intelligent hyperspectral remote sensing image processing based on artificial immune systems, are to solve various problem in hyperspectral remote sensing, such as image classification, spectral matching and decomposition of mixture pixels, by intelligence methods based on artificial immune systems. The proposed algorithm based on multiple-valued immune network is to perform remote sensing image classification. By the experiment, it is demonstrated that our method is superior to traditional algorithms and is an effective option for remote sensing image classification.