Infrared and visible image fusion aims to enhance scene representation by integrating complementary sensor data. However, existing methods fail to reconcile spectral fidelity with structural consistency. For one thing, grayscale fusion approaches preserve structural details by discarding color information, inherently sacrificing spectral fidelity. For another, color fusion techniques maintain spectral authenticity but compromise details and structural consistency due to the misaligned chromatic information. To bridge the gap, we present the Dif-CDFusion, which resolves the conflict between spectral fidelity and the preservation of structural details through diffusion-based feature extraction and common-differential alternating feature fusion. By individually constructing a denoising diffusion process in latent space to model multi-channel spectral distributions, our approach extracts diffusion features that preserve color integrity while capturing complete spectral information for texture retention. Subsequently, we design a common-differential alternate fusion module to alternately integrate differential and common mode components within diffusion features, enhancing both structual details and thermal target salience. Extensive experiments demonstrate that our Dif-CDFusion achieves state-of-the-art performance both quantitatively and qualitatively. The code and datasets are publicly available at https://github.com/ChickenEating/Dif-CDFusion.

Infrared (IR) images inherently face the dual challenges of noise contamination and reduced contrast. However, existing image enhancement methods often overlook the intrinsic correlations between these factors—noise and low contrast—during multistage enhancement processes. Consequently, this oversight leads to a significant reduction in the fidelity of intricate details in IR images. In this article, we present a synergistic IR image enhancement network that simultaneously achieves denoising, contrast improvement, and detail preservation (DCDNet), which breaks down the overall enhancement process into more manageable steps. DCDNet is comprised of a detail awareness unit (DAU), a deep denoising prior (DDP), and a contrast improvement module (CIM). To maintain the details in the IR image, DAU is developed to extract the original detail feature information in DDP and integrate them into the CIM during contrast improvement to improve the final result. The detail information is derived from the encoder of the DDP, which focuses on denoising. The preserved detail features are subsequently incorporated into the decoder of the CIM, which is dedicated to enhancing contrast. Experimental results validate that our proposed approach surpasses other state-of-the-art methods for enhancing IR images in terms of the peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), visual information fidelity (VIF), and performance in downstream tasks. The code and dataset are publicly available athttps://github.com/ChickenEating/IR-Enhancement.

The combination of infrared and visible videos aims to gather more comprehensive feature information from multiple sources and reach superior results on various practical tasks, such as detection and segmentation, over that of a single modality. However, most existing dual-modality object detection algorithms ignore the modal differences and fail to consider the correlation between feature extraction and fusion, which leads to incomplete extraction and inadequate fusion of dual-modality features. Hence, there raises an issue of how to preserve each unique modal feature and fully utilize the complementary infrared and visible information. Facing the above challenges, we propose a novel Differential Feature Awareness Network (DFANet) within antagonistic learning for infrared and visible object detection. The proposed model consists of an Antagonistic Feature Extraction with Divergence (AFED) module used to extract the differential infrared and visible features with unique information, and an Attention-based Differential Feature Fusion (ADFF) module used to fully fuse the extracted differential features. We conduct performance comparisons with existing state-of-the-art models on two benchmark datasets to represent the robustness and superiority of DFANet, and numerous ablation experiments to illustrate its effectiveness.

Pre-disaster information storage is crucial for effective disaster response. The discussion regarding deep learning-based Light Detection and Ranging (Lidar) semantic segmentation technology for indoor small items has been ongoing in recent years. However, the methods applicable to large-scale outdoor Lidar datasets for pre-disaster information storage remain limited. This study aims to propose a novel deep learning-based network for city-scale Lidar semantic segmentation to support pre-disaster information storage, called channel attention and normal-based local feature aggregation network (CNLNet). This network is designed to segment common urban land cover objects, including buildings and vegetation. This network incorporates surface normal information and the channel attention mechanism into the RandLA-Net backbone. Ablation studies have been devised to assess the performance of these two features. During the pre-processing step, color information from optical images is fused with Lidar data. The findings demonstrate that CNLNet can enhance the accuracy of the RandLA-Net backbone by improving mIoU at least 1-2%. Including one of these two features also contributes to the backbone’s improved accuracy. Notably, CNLNet outperforms other well-known networks in terms of accuracy with the test of the public Sementic3D dataset. The study further reveals that the proposed network excels in building segmentation, a crucial facet of pre-disaster information storage. Moreover, the results show that spatial resolution, whether at 0.5m or 10m per pixel for optical images, has limited influence on outcomes. One theoretical contribution of this study is the demonstration of the advantages of integrating either surface normal information or a channel attention mechanism to enhance large-scale outdoor Lidar semantic segmentation. Labeled Lidar datasets have been created for training. The practical contribution is that it can optimize disaster response by efficiently facilitating pre-disaster information storage.

Hazy images captured under ill-posed scenarios with scattering medium (i.e. haze, fog, or smoke) are contaminated in visibility. Inevitably, these images are further degraded by noises owing to real-world imaging. Most existing hazy image enhancement methods perform image dehazing and denoising stage by stage, with the undesirable result that the estimation error of the former stage has to be propagated and amplified in the latter stage, e.g., noise amplification after dehazing. To address this inconsistent degradation, we present an Unsupervised Unified Image Dehazing and Denoising Network, U2D2Net, to remove the haze and suppress the noise simultaneously for a single hazy image. U2D2Net is mainly comprised of an unsupervised dehazing module, an unsupervised denoising module, and a region-similarity fusion strategy. Specifically, we propose an unsupervised transmission-aware dehazing module to restore visibility and suppress depth-dependent noise propagation in the dehazing module. Besides, we design an unsupervised network with a Mean/Max Sub-Sampler in the denoising module. To exploit the correlation and complementary between the previous outputs, a region-similarity fusion strategy is developed to compute the final qualified result. Extensive experiments on both synthetic and real-world datasets illustrate that U2D2Net outperforms other state-of-the-art dehazing and denoising methods in terms of PSNR, SSIM, and subjective visual effects.

Wildfires have attracted considerable attention because of their increasing frequency and severity around the globe. Satellite remote sensing data is a valuable asset for monitoring, and mapping burned areas (BA). However, most global BA products based on optical imagery are limited by cloud coverage and not usable for cloud-prone regions. All-weather Synthetic Aperture Radar (SAR) imagery can be a complement to an optical-based counterpart. In order to exploit the value of phase information of SAR data, this paper aims to propose a framework by developing a visual saliency detection algorithm for BA mapping using Sentinel-1 Interferometric SAR (InSAR) coherence difference image. The results show that the proposed method can effectively improve the coherence difference's accuracy performance. Additionally, we also demonstrate that for C-band Sentinel-1 SAR data, both VV and VH polarized images can be used in BA mapping, but the former would provide slightly better results.

The ancient Aniangzhai (ANZ) landslide in Danba County, Sichuan Province of southwest China was reactivated after a series of complex hazard events that occurred in June 2020. Since then, emergency engineering work was carried out to prevent further failure of the reactivated landslide. This study investigates the multi-temporal optical images (3 m spatial resolution) acquired from the PlanetScope satellite with pixel offset tracking (POT) technique to assess deformation characteristic and spatial-temporal evolution of the reactivated ANZ landslide during the post-failure stage. The relationships between sun illumination differences, temporal baseline of correlation pairs and the uncertainties were explored. The large horizontal displacements over the reactivate ANZ slope were detected from the time-series POT results, showing a significant increase of about 24 m between 24 June 2020 and 11 June 2021. The time series optical POT results revealed that the reactivated ANZ landslide body is gradually slowing down to a steady deformation status since its occurrence in August 2020, indicating the effectiveness of engineering work on the prevention of further landslide.

Most deep learning networks for Lidar semantic segmentation have been devoted to small-scale indoor data and only few of them have focused on large-scale outdoor data. To bridge this gap, this research explores the influences of changing features of deep learning networks on the accuracy of large-scale outdoor Lidar semantic segmentation. Surface normal information and random downsampling layers are the two features considered. Eight scenarios are designed to test them. Point clouds acquired from Kapiti Coast, New Zealand in 2021 with five labeled classes are used for training, validation, and testing stages. Mean intersection over union (mIOU) is the main metric in the validation and test. The findings show that the network adding surface normals with four random downsampling layers whose sampling ratios are 4, 4, 4, and 4 of those layers performs best because of its high mIOU. Moreover, IOU results reflect that the segmentation of buildings performs best between all tested classes.

In 2022, flooding severely violated Australia, resulting in the displacement of residents and damage to property and public facilities. With the rapid development of information technology, it is possible to use Synthetic Aperture Radar (SAR) satellite remote sensing technology and the Quadrotor Unmanned Aerial Vehicle (QUAV) to detect and assess flooding environments. SAR can penetrate the cloud to operate at all times and in all weather, which is ideal for flooding area mapping. However, most SAR-based products are constrained by the flood’s dynamically shifting boundary and spatial and temporal resolution. QUAV is portable and capable of precise positioning despite being ineffective in covering large areas, such as flood-affected areas. Thus, it can complement the SAR counterpart for flood mapping in boundary extraction. This paper aims to propose a framework that mainly fuses satellite SAR and QUAV technology by aggregating the multiple-scale data for double validation and detailing, with enhancement by deep learning-based prediction models and a closed-loop feedback mechanism, to form a novel space-air vertical remote sensing framework. Finally, the selected flood-affected areas in Moama, NSW, Australia, were conducted as a case study. The results show that the proposed method can effectively enhance flood area assessment and mapping.

Early detection of bushfire plays a crucial role in firefighting, fire modelling, and minimising losses of human lives and properties. However, current bushfire monitoring systems have an intrinsic shortcoming because only temperature difference between neighboring pixels is exploited. This paper proposes to also examine a range of other changes occur when a bushfire is ignited, for example, a reduction of vegetation cover, volume scattering of bush and trees, as well as height of vegetation. All of these can be readily measured by optical and radar satellites already in orbits in near real-time, that is, less than two hours after a satellite overpass. Cross-correlation of these measurements has the potential to significantly reduce false alarm of a bushfire, while improving the early detection and measurement of fire spots, and hence make the system much more robust. A case study near Sydney is included here based on Sentinel-1 SAR and Sentinel-2 optical satellite data collected on 10 and 11 October 2020, respectively. This research is a major step forward towards the operational and synergetic use of optical and SAR satellites in bushfire monitoring.

In this paper, the 2020 Aniangzhai Landslide in Danba County in Sichuan province, China was investigated by using multi-temporal SAR and optical datasets. The pre- and post- failure scars of the landslide and debris flow were depicted using high-resolution optical images from the Planetscope satellites. The descending Sentinel-1A/B C-band SAR images were applied to explore the deformation characterization of this event. Advanced time-series InSAR analysis was processed to detect the sliding motion, with identifying the spatial-temporal pattern and evolution of the failure area of the Aniangzhai landslide. The maximum line-of-sight (LOS) deformation rate measured over the slope surface was up to -80 mm/year. Time series analysis of selected measurement points indicates that the cumulative deformation peaked at -113 mm. Most importantly, two significant accelerations were detected at the upper area of the Aningzhai slope before the occurrence of failure. By comparing the time series analysis of points in different sections, it is evidenced that the initial failure of lower part triggered the sliding motion of upper part at the slope.

In this paper, the 2017 Maoxian landslide in Sichuan province, China was investigated by using multi-temporal SAR and optical datasets. The pre- and post-failure scars of the landslide were depicted by using the K-means classification of the Normalized Difference Vegetation Index (NDVI) maps. Two stacks of ascending and descending Sentinel-1A/B C-band SAR images were applied to explore the pre-failure characteristics of this event. Advanced time-series InSAR analysis was processed to detect the pre-failure movements of this event, with identifying the spatial-temporal pattern and evolution of the source area of the Maoxian landslide. The maximum line-of-sight (LOS) deformation rate measured over the slope surface was up to -30 mm/year in the descending track, with only -18 mm/year for the ascending track. Most importantly, an obvious acceleration was detected from the time series analysis of selected measurement points at the source area before the occurrence of failure. By comparing the TS-InSAR result with the precipitation record over this region, it is evidenced that heavy rainfall might be the major triggering factor of the Maoxian Landslide.

This paper addresses the forest height inversion based on the modified RMoG model using repeat-pass Pol-InSAR data. The linear variance of the Gaussian motion distribution in the RMoG model is replaced by a linear standard deviation to describe the volumetric motion heterogeneity and related coherence function is deduced based on that. Furthermore, a forest parameter inversion algorithm is proposed based on this modified model and its performance is investigated with L-band repeat-pass ALOS-1 quad-polarization data acquired over German forest site with temporal baselines of 46 days. Inversion results indicates that in comparison with the traditional RVoG and RMoG methods, the modified method can reduce 27.73% and 8.57% of the overestimation errors caused by the temporal decorrelation.

SqueeSAR, also known as advanced time-series interferometric synthetic aperture radar (ATS-InSAR) method, is a significant improvement of conventional persistent scatterer InSAR (PSInSAR), whereby the concepts of distributed scatterer (DS) and persistent scatterer (PS) are first been introduced, respectively. It is worth noting that during the measurement pixel selection, it is inevitable that a number of PS can be categorized as DS as well, hence resulting in common PS-DS pixels. In order to understand the consistency among these common PS-DS pixels with PSInSAR and ATS-InSAR methods, statistical analyses are conducted with 10 real InSAR image stacks in this letter. The relationship between the goodness-of-fit value and four main factors, including root-mean-square difference, DS percentage, PS-DS/PS ratio, and PS-DS/DS ratio, is studied. It is concluded that Sentinel-1-based TS-InSAR can be less influenced by the goodness-of-fit threshold in comparison with the counterpart result of ALOS-1 under the same parameter setting; finally, conclusions for the threshold settings are given.

The interferometric phases in highly sloped terrain have the characteristics of large fringe density, narrow width, low correlation, and under-sampling. The local fringe frequ- ency (LFF) is a criterion to evaluate the trend and magnitude of the local terrain gradient and can be employed to improve the quality of interferograms. The results of the traditional LFF estimation method can be affected by phase noise, and sometimes the phase unwrapping (PU) operation is also required for some local regions. When it comes to highly sloped terrain, the phenomenon of phase under-sampling may cause incorrectness in the absolute interferometric phase during the operation of PU and may then influence the accuracy of the whole estimation. In order to solve this problem, this letter proposes an extended maximum-likelihood method for LFF estimation based on the multifrequency interferometric synthetic aperture radar (InSAR) data. Through the differences in the LFF between the different frequency InSAR data, the estimation quality map is introduced to modify the large error in certain regions by local 2-D fitting and thus achieves a accurate estimation of LFF in highly sloped terrain. Finally, the estimated results of LFF are used to guide the process of phase filtering. Simulated data and real airborne dual-frequency InSAR data are both employed to validate this proposed method.

In this paper, a creative hybrid adaptive filtering algorithm is designed to combine a modified mode filter with an adaptive median filter for the suppression of interferometric phase noise which is a quality insurance of the following phase unwrapping operation. This combination provides the hybrid filtering algorithm with remarkable flexibility to terrain. At first, this paper presents a modified mode filter with an adaptive shortest sub-interval estimator. Then, it uses the mode filtering result as the variable local phase center of the median filter and improves it with an adaptive filtering window varying with the residue density and correlation coefficient. Finally, a hybrid adaptive filtering algorithm is proposed for the reduction of interferometric phase noise. The studies utilize the simulated Peak data and the real Ayers Rock data as data support. Through analyzing the phase residues and the PSD values of the filtering results, we finally prove the validity of this method.