Deep Learning Approach for Forest Change Detection and Prediction Using Satellite Images

To safeguard the balance of ecology forests play a vital role but unfortunately these forests are under threat dueto many anthropogenic and natural factors. Therefore, watching such forests for any signs of change aimed at their conservation is important. Here, a new method is proposed whereby change detection using satellite imagery from Landsat-8 and the UNET deep learning approach makes for easy identification of forest changes. UNET which has excellent image segmentation capabilities is used to segment images hence it can be applied in detecting deforested areas. This prevents forest cover change detection and classification thus using high-resolution satellite imagery is necessary in this regard. First pre-processing of Landsat-8 images is done. This ensures clarity and relevance of what we are applying. The next step involves training on previously annotated datasets using the UNET model, enabling it to learn different forest conditions. However special indices will be added to improve UNET performance. Additionally, NDVI-normalized difference vegetation index (NDVI) will also be an integral part because it helps with distinguishing different types of vegetation and land use dynamics even though our analysis was limited to time series data collected between 2015–2020 from Periyar forest located at Kerala state in India.