Search

Airborne laser scanning (ALS) point cloud classification is a challenge due to factors including complex scene structure, various densities, surface morphology, and the number of ground objects. A point cloud classification method is presented in this paper, based on content-sensitive multilevel objects (point clusters) in consideration of the density distribution of ground objects. The space projection method is first used to convert the three-dimensional point cloud into a two-dimensional (2D) image. The image is then mapped to the 2D manifold space, and restricted centroidal Voronoi tessellation is built for initial segmentation of content-sensitive point clusters. Thus, the segmentation results take the entity content (density distribution) into account, and the initial classification unit is adapted to the density of ground objects. The normalized cut is then used to segment the initial point clusters to construct content-sensitive multilevel point clusters. Following this, the point-based hierarchical features of each point cluster are extracted, and the multilevel point-cluster feature is constructed by sparse coding and latent Dirichlet allocation models. Finally, the hierarchical classification framework is created based on multilevel point-cluster features, and the AdaBoost classifiers in each level are trained. The recognition results of different levels are combined to effectively improve the classification accuracy of the ALS point cloud in the test process. Two scenes are used to experimentally test the method, and it is compared with three other state-of-the-art techniques.

The use of bi-temporal remote sensing images for detecting changes in land cover is an important means of obtaining surface change information, thus contributing to urban governance and ecological environment monitoring. In this article, we propose a deep learning model named Siam-UNet3+ for high-resolution remote sensing image change detection. This model integrates the full-scale skip connections and full-scale deep supervision of the network UNet3+, which can achieve the multi-scale feature fusion of remote sensing images, effectively avoiding the locality disadvantage of convolution operations. Different from UNet3+, Siam-UNet3+ has made major improvements, including the following: (1) incorporating a Siamese network in the encoder, which can process bi-temporal remote sensing images in parallel; (2) leveraging the residual module as the backbone, which can avoid gradient vanishing (or exploding) and model degradation problems; (3) adding a Triplet Attention module to the decoder, which can avoid information redundancy that may occur in full-scale skip connections and increase the ability to focus on changing patterns; and (4) designing a hybrid loss function consisting of focal loss and dice loss, which is more suitable for remote sensing image change detection tasks. In this study, we conducted change detection experiments using the publicly available LEVIR-CD dataset, as well as two local datasets in Beijing. Through comparative experiments with five other models and ablation experiments, the proposed model Siam-UNet3+ in this article demonstrated significant advantages and improvements in four evaluation metrics, namely, precision, recall, F1-score, and overall accuracy (OA), proving to have great potential in the application to highresolution remote sensing image change detection tasks. [ABSTRACT FROM AUTHOR]