Intelligent Satellite - Based Deforestation Surveillance Using Enhanced Classification

Abstract

Understanding the dynamics of deforestation and land use in adjacent regions is crucial for developing effective forest conservation and management policies. This study presents a novel approach to addressing deforestation by treating it as a multilabel classification (MLC) problem using satellite imagery. We introduce Inception, an advanced model that leverages the self-attention mechanism, thereby eliminating the need for convolution operations typically used in traditional deep learning models for detecting deforestation. Extensive experiments con- ducted on publicly available satellite image datasets demonstrate the effectiveness of Inception in MLC, particularly in scenarios with imbalanced classes. This research marks a significant advancement in utilizing the Inception architecture for deforestation monitoring, emphasizing its potential to enhance the accuracy and sensitivity of land use classification based on satellite images.