New Method Based on Support Vector Machine in Classification for Hyperspectral Data

Cross-validation is a normal method for parameter selection of support vector machine (SVM) which is a novel machine learning method for hyperspectral data classification. Because of the high dimensionality of hyperspectral data, the process of cross-validation will cost more time. For reducing the time of cross-validation and improving classification accuracy, a new combination method of improving sequential minimal optimization (SMO), independent component analysis (ICA) and mixture kernels is proposed. It can be described as follows: first use the improving SMO method to optimize the model of SVM, and then use ICA method to do dimensionality reduction before cross-validation, at last use mixture kernels to do classification of unknown samples. By the experiments, it is proved that this method can guarantee the accuracy of unknown samples classification while reducing the time of cross-validation.