Application of immune network theory for target-oriented multi-spectral remote sensing information mining

To use target information for space transformation in remote sensing data field, artificial immune network theory is introduced to multi-spectral remote sensing information mining, based on the knowledge of target spectrum. First, the target spectrums are fuzzy clustered into several subclasses, to retain different features of target in different subclasses. Then we develop a novel Regional-memory-pattern Artificial Immune Idiotypic Network (RAIN) model based on artificial idiotypic network theory, and train RAIN with subclasses samples. And then, the affinities of the target spectrum and other objects can be calculated according to the immune microscopic dynamics including stimulation and suppression effect. Finally, principal component analysis (PCA) is performed to affinities to explore more weak and hidden information. With its application in Baoguto Area, Xinjiang Uyghur Autonomous Region China, choosing tuffaceous siltstone as target object, the result supports the efficiency of the RAIN-affinity-PCA scheme.