Remote Sensing Based Classification with Feature Fusion Using Machine Learning Techniques

Abstract

There are several uses for remote sensing image scene categorization, which tries to assign semantic categories to remote sensing images based on their contents. The Pass over (POEP) network proposed in this research is a hybrid feature learning as well as end-to-end learning method for remote sensing picture scene interpretation (RSISU). In order to classify scenes using remote sensing, this research suggests integrating feature fusion as well as extraction with classification methods. The newly designed (POEP) has two benefits. First, multi-resolution feature maps created by CNN are integrated using Pass over connections, which has significant advantages for addressing the existence of large-scale variance in RSISU data sets. Here we use UCI dataset with 21 classes of images as database. Initially the image has been pre-processed and by RESNET-50 with Alex net integration of architectures, the features has been extracted. Then by performance analysis and comparative analysis the optimal results are obtained.