COLLABORATIVE REPRESENTATION BASED K-NEAREST NEIGHBOR CLASSIFIER FOR HYPERSPECTRAL IMAGERY

We propose a novel collaborative representation based k-nearest neighbors algorithm for hyperspectral image classification. The proposed method is based on a collaborative representation computed by an l(2)-norm minimization with a Tikhonov regularization matrix. More specific, the testing sample is represented as a linear combination of all the training samples, and the weights for representation are estimated by an l(2)-norm minimization derived closed-form solution. The label of the testing sample is determined by the majority vote of those with k largest representation weights. The experimental results show that the proposed algorithm achieves better performance than several previous algorithms, such as the original k-nearest neighbor classifier and local mean based nearest neighbor classifier.