CLASSIFICATION OF FOREST DEGRADATION CAUSED BY SELECTIVE LOGGING USING DEEP LEARNING TECHNIQUES

Logging in a forest is a controversial topic. When it comes to a dense tropical forest, such as the Amazon, there is difficulty in monitoring illegal logging due to its vast area. Some government and NGO programs focus on this monitoring; however, they track deforestation over large areas, observing the progression of deforestation in large polygons. Monitoring selective logging is a challenge, as large-scale deforestation often begins with tactics to spread out illegal logging activities, removing small portions to make it harder for regulatory agencies to detect. This study aims to explore a dataset from monitored selective logging. A set of cropped images was used to classify segments of regions previously labeled as logging and non-logging areas, followed by training and validation with well-known Deep Neural Network algorithms. Finally, metrics were used to evaluate the classification.