A application of OP-ELM in the remote sensing images recognization

Automatic recognization of remote sensing images is significant. In this article, we use optimally pruned extreme learning machine(OP-ELM) as an automatic classifier to achieve recognization of the objects in remote sensing images based on texture features through grey-level co-occurrence matrix method. OP-ELM is based on the original extreme learning machine(ELM) algorithm with additional steps to make it more robust and genetic. For the recognization, several co-occurrence parameters are computed and compared, and then, we used the more obvious features of the texture of the objects in remote sensing images as criterions. To improve the effection of recognization, we bring in threshold filter, erosion filter and so on after the OP-ELM classification. In the experiments, the results for both computational time and accuracy(Kappa coefficients) are compared to the SVM and BP. As the illustration, OP-ELM performs faster than the other algorithm Without losing the accuracy of the premise.