Individual tree segmentation from point clouds is essential for diverse forest applications. A dual-branch segmentation deep learning network operating at the grid scale was proposed, which includes the semantic segmentation branch for partitioning point clouds of tree trunks and the instance segmentation branch for individual trunk extraction. Meanwhile, the network analyzes input forest points at the grid scale instead of pointwise processing to preserve local geometric information of the forest points while reducing computational load. After extraction of each tree trunk in the understory layer using our network, a hierarchical k-nearest neighbors algorithm based on the extracted trunk parts was employed to accomplish individual tree segmentation. For the forest plots, our proposed approach achieves precision, recall,${F}1$-score, and mean intersection over union (MIoU) of 89.66%, 89.13%, 89.40%, and 90.84%, respectively. These results represent a significant improvement in accuracy and rapid execution capability compared to prior methods.

As desertification is one of the most severe ecological and environmental issues worldwide, monitoring desertification and studying its evolution patterns are highly important for its governance and prevention. In this study, a novel desertification monitoring method is developed that combines the three-dimensional desertification index (TDDI) and Gaussian mixture model (GMM). The results of applying this method to desertification monitoring, which is based on historical Google Earth images, in northern China from 2000 to 2020 indicate that the accuracy of desertification classification using the TDDI and GMM algorithms exceeds 82%. Compared with the national desertification survey statistics, the accuracy of classifying areas with different degrees of desertification exceeds 93.4%. In terms of the stability of the monitoring results under different data source and spatial region conditions, TDDIMODISshows a strong correlation and high consistency with TDDILandsatand TDDIsentinel-2. The overall accuracies are greater than 55%. Additionally, the TDDI comprehensively considers the soil moisture level, vegetation coverage, and surface conditions and reflects the complexity of the desertification process more accurately than the NDVI and DDI.

Mangrove forests are among the most productive of coastal ecosystems, providing a variety of ecological functions and economic value to coastal areas around the world. Accurate identification of mangrove is of great importance for the restoration and conservation of mangrove ecosystems, and for promoting the development of a blue carbon economy and achieving carbon neutral strategies. In this study, a red-edge mangrove index (REMI) was proposed based on Sentinel-2 multispectral images, using red, green, red edge, and SWIR1 bands in the form of a (red edge-red)/(SWIR1-green) combination to highlight the unique green and moisture information of mangrove. Then, the REMI index was combined with the Otsu threshold segmentation algorithm (Otsu) to map the mangrove information in respect of Hainan Island, which has the most abundant mangrove species in China. The results indicate that, when compared with other vegetation indices, such as the normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), mangrove index (MI), normalized difference mangrove index (NDMI), combined mangrove recognition index (CMRI), and mangrove vegetation index (MVI), the REMI showed greater proficiency in distinguishing mangrove from other vegetation. When the REMI was applied to mangrove mapping in Hainan Island, the overall accuracy and kappa coefficient were 95.68% and 0.92, respectively. In addition, the mangrove distribution ranges mapped in this study were compared with existing mangrove products (HGMF_2020 and China National Standard GB/T 7714-2015 (note)), and it was demonstrated that the mangrove distribution ranges identified based on the REMI had high coincidence with the above-mentioned mangrove products. This proves that the REMI has good potential for application in mangrove identification and mapping.

Transient electromagnetic method (TEM) inversion is a complex nonlinear problem with high dimensionality and ill-posedness. Using traditional neural networks based on a gradient algorithm to solve the TEM inversion problem might result in a slow convergence where the model falls easily into local minima. To solve these problems, a wavelet packet denoising (WPD) technology and a regularized extreme learning machine (ELM) algorithm based on the leave-one-out cross-validation (LOO) approach are proposed in this article. First, the WPD method is based on the hard threshold, the Shannon entropy with a Sym15 wavelet, which is provided to suppress the noise from the TEM data. Moreover, the learning process of the ELM model is optimized by randomly setting the hidden layer parameters instead of updating based on the gradient algorithm, which accelerates the learning speed. Finally, the LOO methodology is introduced to optimize the regularization factor of the ELM to improve the prediction accuracy and generalization ability of the approach. The inversion results of two typical TEM layered geoelectric models, two anomalous body models, and one field example are provided to prove the validity and feasibility of the proposed approach. In addition, compared with other traditional methods [ELM, backpropagation (BP), radial basis function (RBF), and linear support vector machine (LSVM)], the introduced method achieves higher inversion accuracy, better stability, and stronger forward data fitting ability, enabling it to effectively solve the TEM inversion problem.

The traditional extreme learning machine (ELM) inversion of transient electromagnetic method (TEM) based on random initial weights is known to be inept for its low-computational efficiency and poor generalization performance. To solve these problems, a kernel principal component online sequential extreme learning machine (KPCOSELM) trained by an improved chaotic differential evolution (ICDE) method is proposed. An additional kernel principal component (KPC) layer is used, which reduces the dimension of TEM data and enhances the computational efficiency of online sequential extreme learning machine (OSELM). Moreover, a novel ICDE algorithm is presented for improving the learning ability and generalization performance of OSELM inversion. In the proposed ICDE, the tent chaotic sequence is adopted to enhance the global exploitation ability, and a constraint factor is added to ensure better convergence. The feasibility and effectiveness of the proposed inversion method are evaluated via four groups of experiments. The inversion results of the synthetic and field examples show that the proposed approach outperformed other methods in terms of computational efficiency and prediction accuracy, and realized satisfactory performance in TEM inversion, which provides a new strategy for the application of neural networks in TEM inversion.

The objective of this paper is to develop a hierarchical classification scheme and propose forest land type precise classification method based on SPOT5, GF-1 images, and other multi-source data, focusing on fine classification of forest land using high resolution remote sensing image in complex mountainous terrain conditions. The experiments were carried out by multi-source data integration, multiple features analysis and multiple classifier combination. The proposed method in this paper have advantages in fine identification of forest land types with high accuracy and high reliability, and the detail degree of fine identification reaches dominant tree species, which could fully meet the needs of forestry applications such as forest resources investigation, forest land change monitoring and thematic map digital update.

Compared with airborne laser scanning, terrestrial laser scanning (TLS) offers ground-based point cloud data of trees and provides greater potential to accurately estimate tree and stand parameters. However, there is a lack of effective methods to accurately identify locations of individual trees from TLS point cloud data. It is also unknown whether the estimation accuracy of the parameters, including tree height (H), diameter at breast height (DBH), and so on, using TLS can meet the requirement of forest management and planning. In this letter, a novel method to effectively process point cloud data and further determine the locations of individual trees in a stand based on the central coordinates of point cloud data on a defined grid according to the largest DBH was developed. Moreover, a point-cloud-data-based convex hull algorithm and the cylinder method were, respectively, used to estimate DBH and H of individual trees. This study was conducted in a pure Chinese fir plantation of 45 trees located in Huang-Feng-Qiao forest farm, You County of Hunan, China. The comparison of the estimated and observed values showed that the obtained tree locations had errors of less than 20 cm, and the relative root mean square errors for the estimates of both DBH and H were less than 5%. This implies that TLS is very promising for the retrieval of tree and stand parameters in forest stands. For the applications of these methods to mixed forests with a structure of multilayer canopies, further examination is needed.