Comparative analysis of computer vision techniques for deforestation detection using LANDSAT 7 satellite images

Abstract

This study explores the application of computer vision techniques for deforestation detection using LANDSAT 7 satellite imagery. Two main classification algorithms, Random Forest and K-Means, were employed to evaluate their effectiveness in identifying deforested areas. Random Forest, a supervised learning method, demonstrated high accuracy and robustness due to its ability to handle labeled data and multiple variables, making it suitable for this task. On the other hand, K-Means, an unsupervised algorithm, struggled with accuracy in this context, highlighting its limitations when dealing with complex real-world data such as satellite imagery. The TOPSIS method was applied for a multi-criteria evaluation of the algorithms, providing a comprehensive comparison that revealed the superiority of Random Forest in this application. TOPSIS was instrumental in objectively assessing the performance of each algorithm based on several criteria, ensuring a balanced and thorough analysis. The study concludes that integrating high-resolution satellite data with sophisticated classification and assessment techniques such as Random Forest and TOPSIS significantly improves deforestation detection and monitoring, contributing to more effective forest conservation strategies.