Edge Computing Environment for Real-Time Automated Satellite Image Classification using Deep Learning Algorithms

Abstract

Abstract Classification of image is one of the decisive processes in image processing. When the classification is coursed up manually, may extend inaccuracy in analysis methods. Consequently, sensing the satellite images leads to more complexity; it has become an important problem unconcealed by technologists and researchers globally. Prevalently, space-to-ground observation technique confining the surface of the earth, civil security and remote sensing for ocean monitoring were included. In order to afford a clear pixel composition a classified image is required. It confers a high computational cost if a high pixel composition of data with a high sustainable feature in order to establish a highly accurate image classification with proper data representation. To ascertain the above mentioned properties we require a better architectural model for finest classification. The righteous delivery of remote sensing images is the upmost utility of satellite images which renders an outcome of less cost effectiveness by a long way. In this paper, we encompass an integrate feature for classifying remote sensing images that is Support Vector Machine with Genetic Algorithm (HGA-SVM) architecture with 3D-CNN and SE-Net. The network Model initiates advents a straightforward and uncomplicated learning model that helps to encompass our proposed architecture that help the model to achieve an even classification of landscapes, a manageable pushover of accuracy and a reasonable computational cost to classify remote sensing images. In this work, we first train the features based on proposed deep learning model in agriculture fields using visible spectrum camera. After that, we develop an AI based edge computing system to fully automate the classification process. For the classification, SVM exemplify with SE-Net based CNN ultimately offshoot an optimal solution and enhanced classification of remote sensing images.