Evaluating Satellite Image Classification: Exploring Methods and Techniques

Abstract

Satellite image classification serves a critical function across various applications, from land cover mapping and urban planning to environmental monitoring and disaster management. In recent years, significant advancements in machine learning and computer vision, coupled with increased accessibility to satellite imagery, have driven considerable progress in this field. Classification techniques for satellite imagery can be primarily divided into three key approaches: automatic, manual, and hybrid. Each approach offers unique advantages but also comes with its own set of limitations. While most methodologies gravitate toward automatic techniques, choosing an appropriate method should be a carefully considered decision based on specific needs. This paper provides an exhaustive review of cutting-edge classification algorithms, including Artificial Neural Networks (ANNs), Classification Trees (CTs), and Support Vector Machines (SVMs). It also offers a comparative analysis between these modern methods and traditional techniques, focusing on their respective performance metrics when applied to satellite data. This study examines key factors affecting remote sensing data classification, including classifier parameter adjustments and combining multiple classifiers. It reviews existing literature to enhance feature selection and classifier optimization for better accuracy. However, it also points out the continuous need for research in image processing to improve classification accuracy.