Fusion Of Hyperspectral Imagery Research Articles

A Joint Convolutional Cross ViT Network for Hyperspectral and Light Detection and Ranging Fusion Classification

The fusion of hyperspectral imagery (HSI) and light detection and ranging (LiDAR) data for classification has received widespread attention and has led to significant progress in research and remote sensing applications. However, existing common CNN architectures suffer from the significant drawback of not being able to model remote sensing images globally, while transformer architectures are not able to capture local features effectively. To address these bottlenecks, this paper proposes a classification framework for multisource remote sensing image fusion. First, a spatial and spectral feature projection network is constructed based on parallel feature extraction by combining HSI and LiDAR data, which is conducive to extracting joint spatial, spectral, and elevation features from different source data. Furthermore, in order to construct local–global nonlinear feature mapping more flexibly, a network architecture coupling together multiscale convolution and a multiscale vision transformer is proposed. Moreover, a plug-and-play nonlocal feature token aggregation module is designed to adaptively adjust the domain offsets between different features, while a class token is employed to reduce the complexity of high-dimensional feature fusion. On three open-source remote sensing datasets, the performance of the proposed multisource fusion classification framework improves about 1% to 3% over other state-of-the-art algorithms.

Classification of planted forest species in southern China with airborne hyperspectral and LiDAR data

ABSTRACT Plantations in southern China are important components of forest ecosystems. However, the distribution of the native tree species is fragmented. Accurate tree species classification can provide scientific support for forest management and measurement. In this study, Gaofeng forest farm in Guangxi Province was selected as the research area, and hyperspectral image combining with LiDAR was used for tree species classification. Derivative reflectance features, vegetation indices, and spatial texture features were extracted from hyperspectral imagery and screened by random forest method. Then, these features were combined with height information provided by LiDAR used for tree species classification, using an object-based classifier. We found that the derivative reflectance feature of hyperspectral imagery was most effective for the tree species with hard leaves, the vegetation index was best suited for the tree species with dark leaves, and the texture feature was ideal for tree species with prominent veins and bright leaves. Tree height data most significantly identified taller tree species. Fusion of hyperspectral imagery with LiDAR data resulted in a maximum classification accuracy of 91.59% and a kappa coefficient of 0.897. The classification accuracy can meet the requirements of forest resource monitoring.

Comparison of bathymetry and seagrass mapping with hyperspectral imagery and airborne bathymetric lidar in a shallow estuarine environment

ABSTRACTWe compared hyperspectral imagery and single-wavelength airborne bathymetric light detection and ranging (lidar) for shallow water (<2 m) bathymetry and seagrass mapping. Both the bathymetric results from hyperspectral imagery and airborne bathymetric lidar reveal that the presence of a strongly reflecting benthic layer under seagrass affects the elevation estimates towards the bottom depth instead of the top of seagrass canopy. Full waveform lidar was also investigated for bathymetry and similar performance to discrete lidar was observed. A provisional classification was performed with limited ground reference samples and four supervised classifiers were applied in the study to investigate the capability of airborne bathymetric lidar and hyperspectral imagery to identify seagrass genera. The overall classification accuracy is highly variable and strongly dependent on the classification strategy used. Features from bathymetric lidar alone are not sufficient for substrate classification, while hyperspectral imagery alone showed significant capability for substrate classification with over 95% overall accuracy. The fusion of hyperspectral imagery and bathymetric lidar only marginally improved the overall accuracy of seagrass classification.

Fusion of waveform LiDAR data and hyperspectral imagery for land cover classification

Current research into the fusion of hyperspectral imagery (HI) and full waveform LiDAR (Light Detection And Ranging) has relied on first processing the full waveform LiDAR (FWL) data to a set of discrete returns before merging because the data structure and sampling interval of HI and FWL are distinctly different. However, additional information about target properties can potentially be recovered if the waveform shape is preserved in the fusion process. This paper proposes a “voxelization” method to register FWL data to HI by dividing the waveform data into voxels, and then synthesizing all waveforms which intersect a voxel column into one three-dimensional superposition waveform: the synthesized waveform (SWF). A vertical energy distribution coefficients (VEDC) feature is proposed for extracting features from SWF, and then the SWF and HI are fused to form a complete feature space for classification. A pairwise classifier was adapted and completed using both Maximum Likelihood and Support Vector Machine classifiers for the combined SWF/HI features. Results show that this method of generating SWF from FWL data can effectively preserve information from the original waveforms, and the fusion of SWF and HI enhanced land cover classification compared to both using either data set alone or the merging of HI with a discrete LiDAR return point cloud.