Remote sensing image classification based on SVM classifier

Abstract

How to choose the kernel function of the SVM classifier and function's parameters affects system's generalization and operating speed directly. It takes Cross Validation and Grid Search to validate the performance of Radial Basis Kernel, Polynomial Kernel and Sigmoid Kernel functions in Multi-class Classification, which can not only deduce the capability of SVM but also prove the effectiveness of Grid Search in finding optimized characteristics. Finally, the three SVM classifier kernel functions are used to classify BSQ remote sensing images in TM6 band, and the experimental data prove their feasibility and high efficiency.