FOREST PLANTATION DETECTION THROUGH DEEP SEMANTIC SEGMENTATION

Abstract. Forest plantations play an important role ecologically, contribute to carbon sequestration and support billions of dollars economic activity each year through sustainable forest management sector value chains. As the global demand for products services increases, marketplace is seeking more reliable data on plantations. Remote sensing technologies allied with machine learning, most recently deep learning techniques, provide valuable inventorying related valuation products. In this work, semantic segmentation U-net architecture was used detect plantation areas using Sentinel-2 CBERS-4A images different Brazil. First, models were built from area Centre-East Paraná State, then best tested in 3 new that present characteristics. The achieved promising results similar ones training set, F1-score ranging 0.9171 0.9499 Kappa values between 0.8712 0.9272, demonstrating feasibility