AL-BIRUNI EARTH RADIUS OPTIMIZATION FOR ENHANCED ENVIRONMENTAL DATA ANALYSIS IN REMOTE SENSING IMAGERY

Abstract

Environmental observation techniques that use remote sensing (RS) scene image categorization play an important role in the widespread use of RS data in both civil and military domains. However, the characteristics of an RS dataset, such as its increased dimensionality and limited number of available labeled examples, present practical and scientific difficulties when attempting RS image classification. As a result of the significant advancements in RS scene image classification with deep transfer learning, there are now numerous opportunities for scientific studies and research. This study focuses on the implementation of the Al-Biruni Earth Radius Optimization with deep transfer learning-based scene image classification (AERODTL-SIC) technique on RS images. The proposed AERODTL-SIC method utilized a deep convolutional neural network-based SqueezeNet method to extract features from RS images in order to determine the various types of scenes. The AERODTL-SIC technique exploits a deep autoencoder neural network (DAENN) for scene image classification. The parameters for the DAENN model were perfectly selected using the AERO model, resulting in improved classification performance. An extensive ranging simulation result was obtained on the benchmark RS image database to demonstrate the efficiency of the AERODTL-SIC algorithm, outperforming current methodologies on a variety of performance metrics.