Hyperspectral remote sensing image classification based on support vector machine

Some traditional algorithms used for hyperspectral remote sensing image classification have some problems such as low computing rate, low accuracy and hard for convergence. According to SVM theory, the classification model based on SVM was constructed. By experimenting with hyperspectral image of 64 bands captured by OMIS sensor, the classification accuracy of SVM using different kernel function was analyzed, and the values of C and γ were gained by grid researching. The results indicate that the radial basis kernel function of SVM has the highest accuracy and it can be well used for hyperspectral remote sensing image classification. SVM classifier has more advantages in the classification in contrast with radial basis function neural network classifier and Minimum Distance Classifier (MDC).