Forest Aerial Image Segmentation Through Satellite Images Using Refine U-Net Model

Satellite imagery is a powerful tool for monitoring the Earth and its changes, including deforestation rates. Deep learning techniques can be used to automate this monitoring process. In this research paper, we propose a Refine-U-Net model for monitoring forest areas using satellite imagery. We employ the U-Net with Refine-Net model for image segmentation to achieve this task. Refine-Net is a deep learning model that can effectively refine the segmentation results while retaining the high-level features of the image. We develop a more efficient and accurate method for monitoring forests. The Refine-U-Net model achieves 99.33% accuracy and 93.05 of mean IoU values. The proposed model can have significant improvement in forest aerial image segmentation over the baselines.