Improved Semisupervised UNet Deep Learning Model for Forest Height Mapping With Satellite SAR and Optical Data

In this study, we introduce an improved semisupervised deep learning approach, and demonstrate its suitability for modeling the relationship between forest structural parameters and satellite remote sensing imagery and producing forest maps. The improved approach is based on a popular UNet model, modified and fine-tuned to improve the forest parameter prediction performance. Within the improved model, squeeze-and-excitation blocks are embedded to recalibrate the multisource features via retrieved channel-wise self-attention and a novel cross-pseudo regression strategy is implemented to train the model in a semisupervised way. The improvement imposes consistency learning on two perturbed network branches: 1) generating regression pseudo-reference; 2) expanding the dataset size. For demonstration, we used satellite synthetic aperture radar (SAR) Sentinel-1 and multispectral optical Sentinel-2 images as remote sensing data, complemented with reference data represented by forest tree height as one of the key forest structural variables. The study area is located in a boreal forestland in Central Finland. Proposed approach showed larger accuracy compared to traditional machine learning methods such as random forests and boosting trees, and baseline UNet model. Best accuracy figures for forest tree height were achieved with combined SAR and optical imagery and were as small as 24.1% root-mean-square error (RMSE) on pixel-level and 15.4% RMSE on forest stand level.