Recently, convolutional neural networks (CNNs) showed excellent performance in many tasks, such as computer vision and remote sensing semantic segmentation. Especially, the ability to learn high-representation features of CNN draws much attention. And random forest (RF) algorithm, on the other hand, is widely applied for variables selection, classification, and regression. Based on the previous fusion models that fused CNN with the other models, such as conditional random fields (CRFs), support vector machine (SVM), and RF, this article tested a method based on the fusion of an RF classifier and the CNN for a very high resolution remote sensing (VHRRS) based forests mapping. The study area is located in the south of China and the main purpose was to precisely distinguish Lei bamboo forests from the other subtropical forests. The main novelties of this article are as follows. First, a test was conducted to confirm if a fusion of CNN and RF make an improvement in the VHRRS information extraction. Second, based on RF, variables with high importance were selected. Then, a test was again conducted to confirm if the learning from the selected variables will further give better results.

Above-ground biomass (AGB) directly relates to the productivity of forests. Precisely, AGB mapping for regional forests based on very high resolution (VHR) imagery is widely needed for evaluation of productivity. However, the diversity of variables and algorithms and the difficulties inherent in high resolution optical imagery make it complex. In this paper, we explored the potentials of the state-of-art algorithm convolutional neural networks (CNNs), which are widely used for its high-level representation, but rarely applied for AGB estimation. Four experiments were carried out to compare the performance of CNNs and other state-of-art Machine Learning (ML) algorithms: (1) performance of CNN using bands, (2) performance of Random Forest (RF), support vector regression (SVR), artificial neural network (ANN) on bands, and vegetation indices (VIs). (3) Performance of RF, SVR, and ANN on gray-level co-occurrence matrices (GLCM), and exploratory spatial data analysis (ESDA), and (4) performance of RF, SVR, and ANN based on all combined data and ESDA+VIs. CNNs reached satisfactory results (with R2 = 0.943) even with limited input variables (i.e., only bands). In comparison, RF and SVR with elaborately designed data obtained slightly better accuracy than CNN. For examples, RF based on GLCM textures reached an R2 of 0.979 and RF based on all combined data reached a close R2 of 0.974. However, the results of ANN were much worse (with the best R2 of 0.885).

No abstract

No abstract

No abstract