Remote Sensing Scene Type Classification Using Multi-Trial Vector-Based Differential Evolution Algorithm and Multi-Support Vector Machine Classifier

Abstract

In recent decades, remote sensing scene type classification becomes a challenging task in remote sensing applications. In this paper, a new model is proposed for multi-class scene type classification in remote sensing images. Firstly, the aerial images are collected from the Aerial Image Dataset (AID), University of California Merced (UC Merced) and REmote Sensing Image Scene Classification 45 (RESISC45) datasets. Next, AlexNet, GoogLeNet, ResNet 18, and Visual Geometric Group (VGG) 19 models are used for extracting feature vectors from the collected aerial images. After feature extraction, the Multi-Trial vector based Differential Evolution (MTDE) algorithm is proposed to choose active feature vectors for better classification and to reduce system complexity and time consumption. The selected active features are fed to the Multi Support Vector Machine (MSVM) for final scene type classification. The simulation results showed that the proposed MTDE-MSVM model obtained high classification accuracy of 99.41%, 99.59% and 99.74% on RESISC45, AID and UC Merced datasets.