Semantic change detection (SCD) aims to recognize land cover transitions from remote sensing images of the given scene acquired at different times. The semantic change maps produced by SCD can provide not only the locations of changes but also the detailed change types (e.g., “from-to” change type). This exhaustive change information plays a significant role in various applications. Postclassification methods with multitemporal remote sensing images have been widely used in SCD. However, many existing methods suffer from the accumulation of misclassification errors. In this article, a deep Siamese postclassification fusion network (PCFN) is proposed to address this problem. PCFN is composed of the Siamese classification network (SCN) for land cover mapping (LCM) and soft fusion network (SFN) for postclassification SCD, respectively. In PCFN, SCN is designed to effectively integrate the temporal correlation between two images by processing the joint features, further improving classification performance. Then, SFN is constructed to determine changes and identify the specific change types through automatic soft fusion. SFN fuses the multitemporal LCM features generated by SCN, then adaptively maps them to the decision space for soft fusion during network training, and predicts the final semantic change maps. Extensive experimental results on two challenging SCD datasets demonstrate that our method can alleviate the error accumulation effectively by combining temporal correlation integration and soft fusion, and achieve promising performance superior to the other state-of-the-art methods in SCD.

Hyperspectral image (HSI) super-resolution generally means the fusion of low-spatial resolution hyperspectral image (LRHSI) and high-spatial resolution multi-spectral (MSI)/panchromatic (PAN) image (HRMPI) to get high-spatial resolution HSI (HRHSI). Existing fusion methods have not sufficiently considered the huge spectral and spatial resolution difference between the LRHSI and HRMPI. In addition, most DL-based methods that adopt the CNN structure are limited by its local feature learning, and it is difficult to exploit the global dependency of image features. To fully adapt to the huge modality difference between LRHSI and HRMPI and release the limitation of local feature learning, we design the cross spectral-scale and shift-window based cross spatial-scale non-local attention networks (CSSNet) to effectively fuse the LRHSI and HRMPI. These two networks could explicitly learn the spectral and spatial correlations between two input images. These correlations are then used to reconstruct the HRHSI feature, which makes the obtained HRHSI feature to maintain the spectral and spatial feature consistency with the input images. Finally, a ‘feature aggregation module’ is designed to aggregate the image features from these two networks and output the fused HRHSI. Extensive experimental results on both HM-fusion (fusion with MSI) and HP-fusion (fusion with PAN image) tasks demonstrate CSSNet’s state-of-the-art (SOTA) performance compared to other fusion methods. (The codes could be available at https://github.com/rs-lsl/CSSNet).

The fusion between the low resolution hyperspectral image (LRHSI) and the panchromatic (PAN) image could obtain the high-resolution hyperspectral image (HRHSI). Recently, deep learning (DL)-based fusion methods have been explored widely due to their powerful feature learning ability. However, most DL-based methods that use the one-step fusion manner can suffer from the blurring effect. In addition, they have not fully utilized the spatial and spectral feature information of two input images, which hinders the improvement of the resulting image quality. Therefore, to fully mitigate the blurring effect and utilize two input images, we propose a dual conditional diffusion models-based fusion network (DCDMF) to obtain the fused HRHSI. The conditional diffusion model (CDM) can generate the high quality image with realistic details in an iterative denoising manner (in the inference sampling stage) other than the one-step fusion manner, which could mitigate the blurring effect greatly. To improve the spatial and spectral fidelity of the fused HRHSI, we propose the dual spatial and spectral CDM (two noise prediction networks with different conditional input) to respectively extract the image feature from the LRHSI and PAN images with different image characteristics and reconstruct the corresponding HRHSI feature and fuse them. In addition, considering the high-dimensional property of the HSI, we pre-train an auto-encoder to encode the HSI into the low-dimensional latent space with more discriminate features to reduce the computational cost. Based on the auto-encoder, we also perform the image generation process in the residual latent space to focus on learning the residual latent spatial details. Extensive experimental results on three datasets show the superiority of our method over several state-of-the-art (SOTA) methods. (The ziyuan dataset and codes could be available at https://github.com/rs-lsl/DCDMF).

Pansharpening means to fuse the low spatial resolution multi-spectral image (LRMSI) and the panchromatic (PAN) image to get the high resolution multi-spectral image (HRMSI). Due to the powerful feature learning ability of the deep-learning (DL), DL-based unsupervised fusion methods have been developed explosively. However, most of the fusion methods are difficult to fully explore and utilize the correct spatial and spectral correlation between the LRMSI, HRMSI and PAN images. In addition, the CNN-dominated fusion framework is limited by its local feature learning without exploring the global feature dependency to further enhance the image feature. Therefore, to fully exploit the correct correlations between LRMSI, HRMSI and PAN images and to explore the global feature dependency, we design a cross-scale unsupervised fusion network (CSFNet). This network is composed of two cross spectral and spatial scale's non-local attention blocks to effectively fuse the LRMSI and PAN image features. And the fusion strategy is implemented by mapping the computed non-local similarity from the low resolution scale to the high resolution scale and outputs the reconstructed HRMSI feature. The experimental results on two datasets show that it achieves state-of-the-art performance compared to other fusion methods.

The hyperspectral image (HSI) has been widely used in many applications due to its fruitful spectral information. However, the limitation of imaging sensors has reduced its spatial resolution that causes detail loss. One solution is to fuse the low spatial resolution hyperspectral image (LR-HSI) and the panchromatic image (PAN) with inverse features to get the high-resolution hyperspectral image (HR-HSI). Most of the existing fusion methods just focus on small fusion ratios like 4 or 6, which might be impractical for some large ratios’ HSI and PAN image pairs. Moreover, the ill-posedness of restoring detail information in HSI with hundreds of bands from PAN image with only one band has not been solved effectively, especially under large fusion ratios. Therefore, a lightweight unmixing-based pan-guided fusion network (Pgnet) is proposed to mitigate this ill-posedness and improve the fusion performance significantly. Note that the fusion process of the proposed network is under the projected low-dimensional abundance subspace with an extremely large fusion ratio of 16. Furthermore, based on the linear and nonlinear relationships between the PAN intensity and abundance, an interpretable PAN detail inject network (PDIN) is designed to inject the PAN details into the abundance feature efficiently. Comprehensive experiments on simulated and real datasets demonstrate the superiority and generality of our method over several state-of-the-art (SOTA) methods qualitatively and quantitatively (The codes in pytorch and paddle versions and dataset could be available at https://github.com/rslsl/Pgnet).

Due to its capacity of long-time automatic observation, the Vaisala CL51 Ceilometer, which is a simple single wavelength lidar, has great potential to retrieve aerosol vertical profiles. However, the backscattering signals from ceilometers around 910 nm, which are seriously affected by background signals and water vapor absorption, strongly limits the performance of aerosol retrievals. To overcome this issue, a signal correction process would be crucial to reduce errors of backscattering signals of the CL51 ceilometer. Herein, we develop a signal correction method including background signal correction and efficient water vapor correction. Using the profile observed by the collocated Raman lidar as reference data, we demonstrate that the signal correction significantly improves the accuracy of aerosol measurements from the ceilometer, reducing the median relative error between the two instruments from 29.34% to 21.54%. Although the median error remains slightly above the generally acceptable level, the improvement is nonetheless evident and meaningful. We further indicate that the water vapor correction, based on the humidity profiles, has effectively scaled the underestimated ceilometer backscatter profiles, particularly under humid conditions. The water correction method is validated in Leipzig, and the results imply the effectiveness of water vapor correction in different locations. For individual profiles, at around 1.3 km, where the largest profile differences occur, the relative error between the original CL61 and PollyXT decreases from 28.0% to 13.2% after water vapor correction. Our findings underscore the importance of the ceilometer in capturing the vertical distribution of aerosols through refined signal processing, offering a practical approach for observing the atmosphere.