Forest Segmentation with U-Net in Satellite Images

Automatic identification of different types of forests holds significant potential for land management applications. For instance, precise identification of forest types could aid in targeted conservation efforts or facilitate sustainable resource utilization. The ability to distinguish between ash and oak forests, for example, may inform land managers about the biodiversity and ecological characteristics of specific areas, guiding more effective conservation strategies. This paper presents a U-Net model trained using WorldView-2 satellite images for such purposes. The dataset is comprised of 8-channel satellite images and masks labeling each pixel as one of 11 forest types. As the size of the reliable, expert-crafted ground truth was insufficient it had to be semi-automatically extended to a larger area to enable training. The resulting synthetic mask was less reliable which posed a significant challenge. The impact of image resolution was also examined by comparing two U-Net models: one trained on full-resolution images, and another on reduced-resolution images. Despite limitations posed by the unreliable ground truth, the results are promising for some classes. Additionally, we found that the accuracy did not significantly deteriorate with lower resolution.