Investigation of Ionospheric Disturbance in ALOS PALSAR Images Using Machine Learning

Abstract

The ionospheric disturbance is a very significant problem in remote sensing and navigation related applications. Dynamic ionospheric layers will witness variation of the plasmatic content which will results in generation of distorted radar images and often the most of the information captured by the image will be lost. The SAR images have vital applications in disaster monitoring, earth’s surface monitoring, mapping of afforestation urbanizations hence the precision level required is higher. In the present study a fusion model of machine and deep learning approach is conducted a combination of GoogLeNet model with KNN, decision tree and SVM classifier is implemented for detection of Synthetic aperture radar image that is subjected to ionospheric variation. The results prove the detection of image affected by ionosphere can be detected put to 96 % and effective in implementing in real time. The data considered in the study is ALOSPALSAR data which is a L band radar data, the raw data is processed in a software called Envi SAR and the images as subjected to preliminary detection phase change method to check image is subjected to ionospheric and latter machine and deep learning approaches are considered.