Abstract：The paper uses multi-spectral image and hyperspectral image of the same time in the same area as the research target, and employs four different kernel functions of SVM classification algorithm to make experiments between these two images based on the premise that the research has the same test samples and identifying samples.The experiments show that for multi-spectral image, RBF kernel function classification will produce the maximum classification precision, while Sigmoid is at its minimum;for hyperspectral images, Linear kernel function will achieve the maximum classification precision, and Sigmoid is at its minimum;for the same resolution remote sensing images at the same area, on the condition of the same classification standard, the classification precision of multi-spectral image is similar to that of hyperspectral image.

Keyword：support vector machine;kernel function;multi-source RS image;