Tropical forests are of vital importance for maintaining biodiversity, regulating climate and material cycles while facing deforestation, agricultural reclamation, and managing various pressures. Remote sensing (RS) can support effective monitoring and mapping approaches for tropical forests, and to facilitate this we propose a deep neural network with an encoder–decoder architecture here to classify tropical forests and their environment. To deal with the complexity of tropical landscapes, this method utilizes a multi-scale convolution neural network (CNN) to expand the receptive field and extract multi-scale features. The model refines the features with several attention modules and fuses them through an upsampling module. A two-stage training strategy is proposed to alleviate misclassifications caused by sample imbalances. A joint loss function based on cross-entropy loss and the generalized Dice loss is applied in the first stage, and the second stage used the focal loss to fine-tune the weights. As a case study, we use Hainan tropical reserves to test the performance of this model. Compared with four state-of-the-art (SOTA) semantic segmentation networks, our network achieves the best performance with two Hainan datasets (mean intersection over union (MIoU) percentages of 85.78% and 82.85%). We also apply the new model to classify a public true color dataset which has 17 semantic classes and obtain results with an 83.75% MIoU. This further demonstrates the applicability and potential of this model in complex classification tasks.

Widely distributed in the vicinity of the equator, tropical forest is one type of forest with the most abundant species worldwide which has a profound effect on global climate change. Therefore, it is of great significance for a country to develop the forest resources inventory and perform dynamic monitoring. Research on the classification of natural forests not only supports the investigation of tropical forests, but also provides the basis for the study of forest species diversity. The dual-polarized SAR data from Sentinel-1A sensor and the optical remote sensing data from Landsat-8 sensor were used for classification of Hainan island tropical natural forest. First, we analyzed the single-band, multi-band, normalized difference vegetation index (NDVI) characteristics of optical data, and the single-phase, multi-temporal backscattering characteristics of SAR data. Then, optical and backscattering characteristics were selected for natural forest classification whereby the natural forest range of Hainan Island was extracted by using support vector machine (SVM). The natural tropical forest was classified into five types: tropical rain forest, tropical monsoon forest, evergreen coniferous forest, deciduous broad-leaved mixed forest and coastal forest. The accuracy of classification results was verified and evaluated based on a combination of field survey data and Hainan forestry survey data. The overall accuracy of the classification exceeded 80%. The results provide a reliable remote sensing classification method for Hainan island tropical forest classification. This dataset also has some reference value for the study of tropical natural forest classification in other areas.