Xuejian Sun

Hyperspectral- multispectral image fusion (HMF) aims to recover a high-resolution hyperspectral image (HrHSI) from an observed high-resolution multispectral image (HrMSI) and a low-resolution hyperspectral image (LrHSI). Traditional HMF methods often suffer from over-reliance on degradation priors. While current blind HMF methods reduce this dependence by estimating the spectral degradation process, they still struggle with common real-world challenges, particularly spatial misregistration and spectral range non-overlap between the observed pair. To address these limitations, we propose a novel blind HMF framework that explicitly tackles both issues. Specifically, we introduce a dedicated image registration strategy before the fusion stage, which explicitly strengthens the spatial mapping relationship between the images and thus eliminates the need for strict initial spatial alignment assumptions. Building upon this, we establish an optimization model for spectral response function (SRF) estimation. The proposed method then employs a decoupled reconstruction mechanism that partitions the HMF process based on spectral band overlap, followed by a spectral recovery module (SRM) to further refine the fused HrHSI, ensuring high spectral accuracy across the entire spectrum. Experimental results demonstrate the exceptional performance of the proposed method in challenging HMF scenarios with significant spatial misregistration. Compared to existing approaches, the proposed method exhibits conclusive superior robustness to spatial misregistration in both SRF estimation and image fusion, achieving higher spectral accuracy in non-overlapping bands and delivering better spatial-spectral reconstruction. Notably, the SRM in the proposed method is highly portable and can be integrated as a general optimization component into other HMF methods to further enhance their performance.

Crop classification is critical for precision agriculture, food security and policymaking. While the fusion of multi-sensor data and deep learning holds significant potential for enhancing crop classification, the application of deep learning-based spectral matching techniques in this domain remains underexplored. To address this gap, this study proposes a novel Time-Aware Siamese Network (TASN) for Spectrotemporal Signature (STS) matching, focusing on kharif crops across three districts in Punjab, Pakistan. The research integrates fusion of Landsat-8/9 and Sentinel-2 data to generate a High-Spatial and High-Temporal (HSHT) dataset, determine optimal temporal lengths for STS extraction using Jeffries-Matusita (JM) distance analysis, applies smoothing techniques to reduce noise in STS, and evaluates TASN’s transfer learning capability for cross-regional crop classification. Results demonstrate that the HSHT dataset significantly enhances temporal resolution, enabling precise crop monitoring. The TASN model achieves 95% accuracy with the Universal Normalized Vegetation Index (UNVI), 92% with NDVI and 91% with EVI. Transfer learning experiments reveal robust adaptability, applying STS of one region to others, TASN attains 98% accuracy for rice classification in Hafizabad using UNVI and 81% to 87% accuracy for multi-crop classification in Lodhran. These findings highlight the efficacy of spectral matching for scalable, transferable crop classification. The study advances agricultural remote seeing by introducing a deep learning framework for STS matching and demonstrating its utility across diverse regions. By combining multi-sensor fusion and TASN’s transferability, this study offers a practical tool for policymakers and farmers, supporting scalable crop monitoring and food security initiatives.

Traditional spatiotemporal fusion methods utilize remote sensing images from two or three adjacent dates to predict missing images. Thus, temporal information in longtime data sets is not fully utilized, increasing the costs of long time series construction in terms of time and labor. Here, we propose a synchronous long time-series completion method using a 3-D fully convolutional neural network (LTSC3D). This method can be used to convert long time-series high-temporal-low-spatial (HTLS)- and high-spatial-low-temporal (HSLT)-resolution remotely sensed images into 4-D data sets. From these data sets, we can then extract and fuse spatiotemporal features to produce synchronous long time-series high-temporal-high-spatial (HTHS)-resolution predictions. The model was tested using simulated and real data sets and compared with four representative traditional spatiotemporal fusion methods. The results demonstrated the high accuracy and high efficiency of our method.

With the growing development of remote sensors, huge volumes of remote sensing data are being utilized in related applications, bringing new challenges to the efficiency and capability of processing huge datasets. Spatiotemporal remote sensing data fusion can restore high spatial and high temporal resolution remote sensing data from multiple remote sensing datasets. However, the current methods require long computing times and are of low efficiency, especially the newly proposed deep learning-based methods. Here, we propose a fast three-dimensional convolutional neural network-based spatiotemporal fusion method (STF3DCNN) using a spatial-temporal-spectral dataset. This method is able to fuse low-spatial high-temporal resolution data (HTLS) and high-spatial low-temporal resolution data (HSLT) in a four-dimensional spatial-temporal-spectral dataset with increasing efficiency, while simultaneously ensuring accuracy. The method was tested using three datasets, and discussions of the network parameters were conducted. In addition, this method was compared with commonly used spatiotemporal fusion methods to verify our conclusion.

Hyperspectral remote sensing plays an important role in a wide variety of fields. However, its specific application for land surface analysis has been constrained due to the different shapes of thick, opaque cloud cover. The reconstruction of missing information obscured by clouds in remote-sensing images is an area of active research. However, most of the available cloud-removal methods are not suitable for hyperspectral images, because they lose the spectral information which is very important for hyperspectral analysis. In this article, we developed a new spectral resolution enhancement method for cloud removal (SREM-CR) from hyperspectral images, with the help of an auxiliary cloud-free multispectral image acquired at different times. In the fixed hyperspectral image, spectra of the cloud cover pixels are reconstructed depending on the relationship between the original hyperspectral and multispectral images. The final resulting image has the same spectral resolution as the original hyperspectral image but without clouds. This approach was tested on two experiments, in which the results were compared by visual interpretation and statistical indices. Our method demonstrated good performance.

The present study utilized time-series Landsat images to explore the spatiotemporal dynamics of urbanization and land use/land-cover (LULC) change in the Kaski District of Nepal from 1988 to 2016. For the specific overtime analysis of change, the LULC transition was clustered into six time periods: 1988–1996, 1996–2000, 2000–2004, 2004–2008, 2008–2013, and 2013–2016. The classification was carried out using a support vector machine (SVM) algorithm and 11 LULC categories were identified. The classified images were further used to predict LULC change scenarios for 2025 and 2035 using the hybrid cellular automata Markov chain (CA-Markov) model. Major hazard risk areas were identified using available databases, satellite images, literature surveys, and field observations. Extensive field visits were carried out for ground truth data acquisition to verify the LULC maps and identify multihazard risk areas. The overall classification accuracy of the LULC map for each year was observed to be from 85% to 93%. We explored the remarkable increase in urban/built-up areas from 24.06 km² in 1988 to 60.74 km² by 2016. A majority of urban/built-up areas were sourced from cultivated land. For the six time periods, totals of 91.04%, 78.68%, 75.90%, 90.44%, 92.35%, and 99.46% of the newly expanded urban land were sourced from cultivated land. Various settlements within and away from the city of Pokhara and cultivated land at the river banks were found at risk. A fragile geological setting, unstable slopes, high precipitation, dense settlement, rampant urbanization, and discrete LULC change are primarily accountable for the increased susceptibility to hazards. The predicted results showed that the urban area is likely to continue to grow by 2025 and 2035. Despite the significant transformation of LULC and the prevalence of multiple hazards, no previous studies have undertaken a long-term time-series and simulation of the LULC scenario. Updated district-level databases of urbanization and hazards related to the Kaski District were lacking. Hence, the research results will assist future researchers and planners in developing sustainable expansion policies that may ensure disaster-resilient sustainable urban development of the study area.

An increasingly common requirement in remote sensing is the integration of hyperspectral data collected simultaneously from different sensors (and fore-optics) operating across different wavelength ranges. Data from one module are often relied on to correct information in the other, such as aerosol optical thickness (AOT) and columnar water vapor (CWV). This paper describes problems associated with this process and recommends an improved strategy for processing remote sensing data, collected from both visible to near-infrared and shortwave infrared modules, to retrieve accurate AOT, CWV, and surface reflectance values. This strategy includes a workflow for radiometric and spatial cross-calibration and a method to retrieve atmospheric parameters and surface reflectance based on a radiative transfer function. This method was tested using data collected with the Compact Airborne Spectrographic Imager (CASI) and SWIR Airborne Spectrographic Imager (SASI) from a site in Huailai County, Hebei Province, China. Various methods for retrieving AOT and CWV specific to this region were assessed. The results showed that retrieving AOT from the remote sensing data required establishing empirical relationships between 465.6 nm/659 nm and 2105 nm, augmented by ground-based reflectance validation data, and minimizing the merit function based on AOT@550 nm optimization. The paper also extends the second-order difference algorithm (SODA) method using Powell’s methods to optimize CWV retrieval. The resulting CWV image has fewer residual surface features compared with the standard methods. The derived remote sensing surface reflectance correlated significantly with the ground spectra of comparable vegetation, cement road and soil targets. Therefore, the method proposed in this paper is reliable enough for integrated atmospheric correction and surface reflectance retrieval from hyperspectral remote sensing data. This study provides a good reference for surface reflectance inversion that lacks synchronized atmospheric parameters.