Kmeans-CM Algorithm With Spectral Angle Mapper for Hyperspectral Image Classification

Hyperspectral images possess the characteristics of high dimensionality, which causes “dimensional disaster” and low classification accuracy, in response to problems, based on traditional k-means algorithm considering importance different bands for classification, also combining both intra-class inter-class information, a Kmeans-CM (K-means with correlation coefficient maximize distance) spectral angle mapper hyperspectral image is proposed. First, weights are defined by introducing variation mapping, measures classification. Second, order intensify between pixels same category, introduced reset distance. Then, information clustering maximizing distance class centers global center reduce effect local optimum effect. Finally, K-means objective function redefined according band intra-class, solving optimally. The overall accuracy method reached 84.47%, 90.08% 80.45% classical data sets Pavia University, Salinas Botswana respectively, results show that proposed owns good performance.