Hyperspectral image (HSI) provides rich spatial and spectral information of ground objects, while synthetic aperture radar (SAR) records scattering information such as shape and structure. Fusion of HSI and SAR can improve the classification performance of land covers. In recent years, graph convolutional networks (GCN) have been widely used in the field of remote sensing due to its advantages in processing non-Euclidean structures, capturing local and global information. In this paper, we propose a spectral-spatial graph convolutional network (SSGCN) for fusion of HSI and SAR. First, the GCN is utilized to extract the spatial information of HSI and SAR. Then, a convolutional neural network is applied to extract the spectral information of HSI. Finally, the extracted spectral and spatial features are merged together followed by a fully connected layer to obtain the final classification result. Experiments on two datasets, i.e., Berlin and Augsburg, reveal that the proposed SSGCN significantly outperforms other representative methods.

Hyperspectral change detection (HCD) aims to identify the changed and unchanged pixels in bitemporal images, which has been applied in various aspects. Currently, many deep learning-based change detection methods have been developed. However, existing change detection methods only focus on changed or temporal information while neglecting the complementary information between them. To solve this issue, a novel cross-temporal symmetric feature fusion network (CTSFFNet) is proposed for change detection of hyperspectral images. First, we perform pixel-wise subtraction and concatenation on the multi-temporal hyperspectral images to obtain the difference data and temporal data, respectively. Then, a three-layer convolutional neural network is performed on the difference data and temporal data to yield the difference and temporal features. Finally, a cross-temporal symmetric feature fusion (CTSSF) module is designed to merge the extracted features followed by a fully connected layer to obtain the final change regions. Experiments on two popular datasets demonstrate that the proposed CTSSFNet achieves superior detection performance compared to other state-of-the-art methods.

Point cloud registration is a prominent topic in computer vision research, with applications including target identification, 3D reconstruction, SLAM, and others. The capture of key points and matching point pairs has evolved into a critical technology for point cloud registration. In this paper, we propose an approach that extracts key points using the Intrinsic Shape Signature (ISS) algorithm. Additionally, we combine the Fast Point Feature Histograms (FPFH) descriptor with curvature to obtain matching point pairs, thereby establishing a solid foundation for point cloud registration. The experiment shows that, compared to the 3D-SIFT technique, the key points derived by the ISS algorithm have a more uniform distribution. Moreover, our proposed combination of the FPFH descriptor with curvature extracts matching point pairs more effectively than using the FPFH descriptor alone.

Hyperspectral and synthetic aperture radar (SAR) image classification, aiming to merge multisource information to boost the precision and reliability of land cover classification, has gained increasing attention. Nevertheless, current techniques still exhibit certain limitations in extracting discriminative features and integrating heterogeneous features. In this work, a graph convolutional fusion network (GCFNet) is proposed for hyperspectral and SAR image classification. First, a spectral residual neural network is employed to extract the spectrum information. Then, a dual-branch graph convolutional network (GCN) is developed to extract the spatial information from hyperspectral and SAR images. Finally, a cross-contextual transformer fusion module is created to merge the spectral and spatial information followed by a dense layer to yield the final prediction outcome. To confirm the performance of the GCFNet, experiments on three datasets (e.g., Berlin, Augsburg, and Yellow River) demonstrate that the GCFNet significantly surpasses other representative methods.

Hyperspectral images (HSIs) offer detailed and abundant spectral-spatial information, holding great potential for ground object change detection (CD). Currently, numerous deep learning-based hyperspectral CD (HCD) models have been studied. However, these methods only consider single difference information or temporal information while neglecting the complementary advantage between difference and temporal information. To solve this issue, this work proposes a multitemporal symmetric fusion network (MTSFNet) for HCD, which involves three steps. First, the difference and temporal data are calculated by a subtraction operation and concatenation operation. Then, a dual-branch convolutional network is developed to capture the difference and temporal features. Finally, a symmetric feature fusion scheme is designed to integrate the extracted features followed by a fully connected layer to derive the detection results. Experiments conducted on multiple well-known datasets reveal that the proposed MTSFNet outperforms other advanced CD approaches in terms of qualitative and quantitative results.

Hyperspectral change detection (HCD) is widely applied in various domains, such as accurate agriculture, disaster assessment, land use, and environmental monitoring. Most of hyperspectral change detection methods aim at extracting and classifying the spectral variation features with dimension reduction and machine learning methods. Different from previous work, this paper proposes an edge-guided hyperspectral change detection method. Specifically, a subtraction operation is adopted to extract difference hyperspectral image. Then, the edge-preserving filtering is performed on the difference HSI so as to extract spectral-spatial features. Next, the number of the extracted features is diminished through the kernel principal component analysis. Finally, the fused features are input into a spectral classifier followed by the edge-preserving filtering to obtain the final change detection result. Experiments on several HCD datasets demonstrate that the proposed method can consistently outperform other advanced approaches in both subjective and objective evaluations when only a limited number of labeled samples are available.

As one of the most important components of plants, the reconstruction of high-precision leaf models is a critical step for building tree models. According to the morphological structure of leaves, this paper proposes a leaf reconstruction method based on the laser point cloud data. The moving least squares method is used to fit the leaf surface to extract the complex profile information of the leaf, and the delaunay triangulation algorithm is used to reconstruct the three-dimensional model of the fitted leaf point cloud data. The results show that the proposed method can not only realize the three-dimensional reconstruction of the leaf, but also reflect the morphological characteristics of real leaf.

Forest resources are of great significance in regulating climate, maintaining biodiversity, and providing ecological products. Accurate identification of tree species is the basis for research and utilization of forest resources. This study combined the characteristics of multi-source data, based on the AISA EAGLE II hyperspectral images and airborne LiDAR point clouds which were obtained in August, 2016. Point cloud characteristics, spectral and texture characteristics were extracted from both datasets. Then SVM was used to classify the main tree species of Genhe experimental area. The results showed that tree species classification accuracy can be improved by using airborne LiDAR and hyperspectral image features.

The boundary feature is of great significance in describing the object shape and constructing 3D model of object. Accurate extraction of boundary features is important to the visualization of the object. This paper presented an accurate extraction algorithm of the indoor boundary features based on point cloud data. The amount of the data was downsampled by the voxelgrid filter. The boundary features of the indoor were extracted by using the angle criterion based on the normal vector and the statistical filtering algorithm. The results show that the boundary features of the indoor can be accurately extracted by the proposed method.