Texture image classification based on support vector machine and bat algorithm

Texture is the vital feature for remote sensing image classification, however, it is hard to be described and recognized by computer vision. As a result, lots of approaches have been presented to identify texture image. Among these methods, support vector machine (SVM) is the most successfully used one, which takes advantages of avoiding local optimum, conquering dimension disaster with small samples. Nevertheless, the selection of the kernel function parameter and error penalty factor has impact on the precision of SVM notably. Some methods have been put forward to learn good parameters for SVM. However, the traditional tuning methods may be inefficient or not robust. Hence, a novel meta-heuristic-bat algorithm is suggested to acquire the optimal parameters for SVM in the paper. In final, experimental results on actual remote sensing texture images manifest that the proposed approach is robust, it is able to distinguish different texture images with high accuracy.