ISPRS Journal of Photogrammetry and Remote Sensing最新文献

Hyperspectral (HS) satellites like PRISMA (PRecursore IperSpettrale della Missione Applicativa) offer remarkable capabilities, yet they are constrained by a relatively coarse spatial resolution, curbing their efficacy in those applications that require pinpoint accuracy. Here we propose a fusion process, aimed at the enhancement of PRISMA HS spatial resolution by using the spatial and spectral information of Sentinel-2 multispectral (MS) data (HS-MS fusion process), validated against four airborne HS flights simultaneous to satellite overpasses on different land use distributions. Adopting the PRISMA panchromatic (PAN) image, the proposed solution was also compared with the results of a HS-PAN pansharpening process. A two-steps operational workflow is proposed, based on two state-of-the-art and open-source algorithms. The first step consisted of the geocoding of PRISMA L2 products using Senintel-2 as reference and was accomplished with the phase-based algorithm implemented in AROSICS (Automated and Robust Open-Source Image Co-registration Software). The geometric displacement in L2 data was found to be between 80 m and 250 m, irregularly spatially distributed throughout the same scene and among scenes, and it was corrected by means of thousands of regularly spatially distributed tie points. A second-order polynomial transformation function was integrated in the algorithm. The second step consisted of employing the HySure (HS Super resolution) fusion algorithm to perform both the HS-MS fusion and the HS-PAN pansharpening, returning a PRISMA HS improved dataset with a spatial resolution of 10 m and 5 m, respectively. Four different per-band accuracy metrics were used to evaluate the accuracy of both products against airborne data. Overall, HS-MS data achieved increased accuracy in all validation metrics, i.e. + 28 % (root mean square error, RMSE), +23 % (spectral angle mapper, SAM), +7% (peak signal-to-noise ratio, PSNR) and + 11 % (universal image quality index, UIQI), with respect of HS-PAN data. These outcomes showed that using the spectral information of Sentinel-2 both spectral and spatial patterns were reconstructed more consistently in three different urban and rural scenarios, avoiding the presence of blur and at-edge artefacts as opposed to HS-PAN pansharpening, therefore suggesting an optimal strategy for satellite HS data resolution enhancement.

The vision enhancement of autonomous underwater vehicle (AUV) has received increasing attention and rapid development in recent years. However, existing methods based on prior knowledge struggle to adapt to all scenarios, while learning-based approaches lack paired datasets from real-world scenes, limiting their enhancement capabilities. Consequently, this severely hampers their generalization and application in AUVs. Besides, the existing deep learning-based methods largely overlook the advantages of prior knowledge-based approaches. To address the aforementioned issues, a novel architecture called CodeUNet is proposed in this paper. Instead of relying on physical scattering models, a real-world scene vision enhancement network based on a codebook prior is considered. First, the VQGAN is pretrained on underwater datasets to obtain a discrete codebook, encapsulating the underwater priors (UPs). The decoder is equipped with a novel feature alignment module that effectively leverages underwater features to generate clean results. Then, the distance between the features and the matches is recalibrated by controllable matching operations, enabling better matching. Extensive experiments demonstrate that CodeUNet outperforms state-of-the-art methods in terms of visual quality and quantitative metrics. The testing results of geometric rotation, SIFT salient point detection, and edge detection applications are shown in this paper, providing strong evidence for the feasibility of CodeUNet in the field of autonomous underwater vehicles. Specifically, on the full reference dataset, the proposed method outperforms most of the 14 state-of-the-art methods in four evaluation metrics, with an improvement of up to 3.7722 compared to MLLE. On the no-reference dataset, the proposed method achieves excellent results, with an improvement of up to 0.0362 compared to MLLE. Links to the dataset and code for this project can be found at: https://github.com/An-Shunmin/CodeUNet.

The Ice, Cloud and Land Elevation Satellite-2 (ICESat-2) is an active spaceborne remote sensing system that utilizes photon-counting LiDAR to capture highly detailed information about under-vegetation terrain and forest structure over vast spatial regions. It facilitates the accurate retrieval of terrain elevation and canopy height information, critical for assessing the global carbon budget and understanding the role of forests in climate change mitigation. However, challenges arise from the characteristics of the ICESat-2 photon-counting LiDAR data, such as their linear distribution, extensive spatial coverage, and substantial residual noise. These challenges hinder the performances of the state-of-the-art methods when applied on ICESat-2 data for extracting ground or top of canopy, while they perform well on airborne LiDAR that is featured with planar distribution, small coverage, and high signal-to-noise ratio. Consequently, this study proposes a novel algorithm termed Adaptive Linear Cloth Simulation Filtering (ALCSF), for the automated extraction of ground and top-of-canopy photons from ICESat-2 signal photons. The ALCSF algorithm innovatively introduces a cloth strip model as a reference to accommodate the distribution characteristics of ICESat-2 photons. Additionally, it employs a terrain-adaptive strategy to adjust the rigidity of the cloth strip by utilizing terrain slope information, thus making ALCSF applicable to large-scale areas with significant topographical changes. Furthermore, the proposed ALCSF addresses noise interference by simultaneously considering the movability of particles of the cloth strip model and the photon distribution during iterative adjustments of the cloth strip. The performance of the ALCSF is evaluated by comparing it with the ICESat-2 Land–Vegetation Along-Track Products (ATL08) across twelve datasets that encompass various times of day and scenes. In the results, the ALCSF exhibits notable improvements over ATL08 products, effectively reducing the root mean square error (RMSE) of ground elevation by 21.8% and canopy height by 25.8%, with superior performance in preserving terrain details. This highlights the significance of ALCSF as a valuable tool for enhancing the accuracy of ICESat-2 land and vegetation products, ultimately contributing to the estimation of the global carbon budget in future studies.

Kinematic laser scanning is a widely-used surveying technique based on light detection and ranging (LiDAR) that enables efficient data acquisition by mounting the laser scanner on a moving platform. In order to obtain a georeferenced point cloud, the trajectory of the moving platform must be accurately known. To this end, most commercial laser scanning systems comprise an inertial measurement unit (IMU) and a global navigation satellite system (GNSS) receiver and antenna. Trajectory estimation is then the task of determining the platform’s position and orientation by integrating measurements from the IMU, GNSS, and possibly the laser scanner itself. Here, we present a comprehensive approach to trajectory estimation for kinematic laser scanning, based on batch least-squares adjustment incorporating pre-processed GNSS positions, raw IMU data and plane-based LiDAR correspondences in a single estimation procedure. In comparison to the classic workflow of Kalman filtering followed by strip adjustment, this is a holistic approach with tight coupling of IMU and LiDAR. For the latter, we extend the data-derived stochastic model for the LiDAR plane observations with prior knowledge of the LiDAR measurement process. The proposed trajectory estimation approach is flexible and allows different system configurations as well as joint registration of multiple independent kinematic datasets. This is demonstrated using as a practical example a combined dataset consisting of two independent data acquisitions from crewed aircraft and uncrewed aerial vehicle. All measurements from both datasets are jointly adjusted in order to obtain a single high-quality point cloud, without the need for ground control. The performance of this approach is evaluated in terms of point cloud consistency, precision, and accuracy. The latter is done by comparison to terrestrially surveyed reference data on the ground. The results show improved consistency, accuracy, and precision compared to a standard workflow, with the RMSE reduced from 7.43 cm to 3.85 cm w.r.t. the reference data surfaces, and the point-to-plane standard deviation on the surfaces reduced from 3.01 cm to 2.44 cm. Although a direct comparison to the state-of-the-art can only be made with caution, we can state that the suggested method performs better in terms of point cloud consistency and precision, while at the same time achieving better absolute accuracy.

State-of-the-art models in semantic segmentation primarily operate on single, static images, generating corresponding segmentation masks. This one-shot approach leaves little room for error correction, as the models lack the capability to integrate multiple observations for enhanced accuracy. Inspired by work on semantic change detection, we address this limitation by introducing a methodology that leverages a sequence of observables generated for each static input image. By adding this “temporal” dimension, we exploit strong signal correlations between successive observations in the sequence to reduce error rates. Our framework, dubbed SSG2 (Semantic Segmentation Generation 2), employs a dual-encoder, single-decoder base network augmented with a sequence model. The base model learns to predict the set intersection, union, and difference of labels from dual-input images. Given a fixed target input image and a set of support images, the sequence model builds the predicted mask of the target by synthesizing the partial views from each sequence step and filtering out noise. We evaluate SSG2 across four diverse datasets: UrbanMonitor, featuring orthoimage tiles from Darwin, Australia with four spectral bands at 0.2 m spatial resolution and a surface model; ISPRS Potsdam, which includes true orthophoto images with multiple spectral bands and a 5 cm ground sampling distance; ISPRS Vahingen, which also includes true orthophoto images and a 9 cm ground sampling distance; and ISIC2018, a medical dataset focused on skin lesion segmentation, particularly melanoma. The SSG2 model demonstrates rapid convergence within the first few tens of epochs and significantly outperforms UNet-like baseline models with the same number of gradient updates. However, the addition of the temporal dimension results in an increased memory footprint. While this could be a limitation, it is offset by the advent of higher-memory GPUs and coding optimizations. Our code is available at https://github.com/feevos/ssg2.

In remote sensing (RS) image change detection (CD) task, many existing CD methods focus more on how to improve accuracy, but they usually have more parameters, higher computational costs, and heavier memory usage. Designing lightweight and performance-sustainable CD model that is more compatible with real-world applications is an urgent problem to be solved. Therefore, we propose a lightweight change detection network, called as robust feature aggregation network (RFANet). To improve representative capability of weaker features extracted from lightweight backbone, a feature reinforcement module (FRM) is proposed. FRM allows current level feature to densely interact and fuse with other level features, thus accomplishing the complementarity of fine-grained details and semantic information. Considering massive objects with rich correlations in RS images, we design semantic split-aggregation module (SSAM) to better capture global semantic information of changed objects. Besides, we present a lightweight decoder containing channel interaction module (CIM), which allows multi-level refined difference features to emphasize changed areas and suppress background and pseudo-changes. Extensive experiments carried out on four challenging RS image CD datasets illustrate that RFANet achieves competitive performance with fewer parameters and lower computational costs. The source code is available at https://github.com/Youzhihui/RFANet.

Long-term, accurate, stable, and continuous aerosol records from space are a major requirement for climate and atmospheric environment research. Due to the limited period span of the single satellite mission, solving the problem requires the combination of multiple satellite missions. In this study, a novel aerosol retrieval algorithm, named enhanced Land General Aerosol Retrieval (e-LaGA), for Visible/Infrared Imager Radiometer Suite (VIIRS) was developed to continue Moderate-resolution Imaging Spectro-radiometer (MODIS) aerosol retrieval. e-LaGA utilizes a Surface Reflectance (SR) relationship model map to more accurately estimate the SR parameter, optimizes the priori aerosol-type map, and improves the multi-band retrieval strategy using the residual-interpolation approach. Additionally, e-LaGA expands the retrieval ability of the traditional Dark Target (DT) algorithm over bright surfaces, and can more accurately retrieve Aerosol Optical Depth (AOD), Fine-mode AOD (AODF), and Fraction (FMF). The validation using global 533 AERONET sites shows that the volume of matchups of AOD (AODF) is approximately 80000, the correlation coefficient (R) is 0.902 (0.879) and the fraction of meeting the expected error envelope (±0.05 ± 0.15τ) is 0.806 (0.804). The inter-comparison indicates that the accuracy of e-LaGA AOD retrievals is comparable to seven commonly used AOD products. The validation performance and spatial distribution pattern of e-LaGA AODF and FMF are in good agreement with the POLDER (Polarization and Directionality of the Earth's Reflectances) products. e-LaGA is also applied to MODIS sensors. The VIIRS e-LaGA AOD, AODF, and FMF retrievals have high consistency with MODIS e-LaGA. Their average bias of AOD is 0.006, which is smaller than 0.024 for the Deep Blue and 0.011 for the DT. These preliminary results demonstrate the robustness of the e-LaGA algorithm and its potential for establishing a long-term climate record of total and fine-mode aerosol by combining multiple satellite missions, which is expected to reduce the uncertainty of climate change research.

Deep learning models centered on retrieval have made significant strides in point cloud place recognition. However, existing approaches struggle to generate discriminative global descriptors and often rely on labor-intensive negative sample mining. Such constraints limit their usability in dynamic and open-world scenarios. To address these challenges, we introduce LAWS, a pioneering classification-centric neural framework that emphasizes looking at the whole scene for superior point cloud descriptor extraction. Central to our approach is the space partitioning design, constructed to provide holistic scene supervision, ensuring the comprehensive learning of scene features. To counteract potential ambiguities arising from the single orthogonal partition boundary, a complementary mechanism of repartitioning space diagonally is specifically designed to dispel classification uncertainties. Under the enhanced partitioning mechanism, the space is separated into several classes and groups. Furthermore, to prevent knowledge forgetting between different groups, a special training strategy is employed, allowing for distinct training of each group. The extensive experiments, encompassing both indoor and outdoor settings and different tasks, validate the generality of LAWS. It not only outperforms contemporary methods but also demonstrates a profound generalization ability across various unseen environments and sensor modalities. Our method achieves a 2.6% higher average top-1 recall on Oxford RobotCar Dataset and a 7.8% higher average recall when generalized to In-house Dataset compared with retrieval-based methods. Furthermore, LAWS also outperforms retrieval-based methods in terms of $F_1$ score, with improvements of 12.7 and 29.2 on the MulRan and KITTI datasets, respectively. Notably, the average localization accuracy of LAWS in indoor environments reached about 68.1%. Moreover, the scalability and efficiency places LAWS in a leading position for continuous exploration and long-term autonomy. Our code is available at https://github.com/BrusonX/LAWS.

The wide application of urban outdoor surveillance systems has greatly improved the efficiency of urban management and social security index. However, most of the existing urban outdoor surveillance cameras lack the records of important parameters such as geospatial coordinates, field of view angle and lens distortion, which brings difficulties to the unified management and layout optimization of the cameras, geospatial analysis of video data, and the computer vision applications such as the trajectory tracking of moving targets. To address this problem, this paper designs a marker with a chessboard pattern and a positioning device, makes the marker move in outdoor space through vehicles and other mobile carriers, and utilizes the marker image captured by the surveillance camera and the spatial position information obtained by the positioning device to batch calibrate the outdoor surveillance cameras and calculate its geospatial coordinates and field of view angle, which achieves the rapid acquisition of important parameters of the surveillance camera, and provides a new method for the rapid calibration of urban outdoor surveillance cameras, which contributes to the informationization management of urban surveillance resources and the spatial analysis and computation of surveillance video data, and make it play a greater role in the application of smart transportation and smart city. Taking the outdoor surveillance cameras within 2.5Km² of a city as an example, calibration tests were performed on 295 surveillance cameras in the test area, and the geospatial coordinates, field of view angle and lens parameters of 269 surveillance cameras were obtained, and the average error of the spatial position was 0.527 m, and the maximum error was 1.573 m, and the average error of the field of view angle was 1.63°, and the maximum error was 3.4°, which verified the effectiveness and accuracy of the method in this paper.

Knowledge distillation (KD) has been one of the most potential methods to implement a lightweight detector, which plays a significant role in satellite in-orbit processing and unmanned aerial vehicle tracking. However, existing distillation paradigms exhibit limited accuracy in detecting arbitrary-oriented objects represented with rotated bounding boxes in remote sensing images. This issue is attributed to two aspects: (i) boundary discontinuity localization distillation, caused by angle periodicity of rotated bounding boxes, and (ii) spatial ossified feature distillation, induced by orientation-agnostic knowledge transitive regions, both of which contribute to ambiguous orientation estimation of objects. To address these issues, we propose an effective KD method called Orientation Distillation (OD) via anti-ambiguous spatial transformation, which consists of two modules. (i) Anti-ambiguous Location Prediction (ALP) module reformulates the regression transformation between teacher–student bounding boxes as Gaussian distributions fitting procedure. These distributions with distilled potential are optimized to accurately localize objects with the aid of boundary continuity cost. (ii) Orientation-guided Feature Calibration (OFC) module employs a learnable affine matrix to augment fixed CNN sampling grid into a spatially remapped one, which bridges between the multi-scale feature of teacher and student for effectively delivering the refined oriented awareness within adaptively distillation regions. Overall, OD customizes the spatial transformation of bounding box representation and sampling grid to transfer anti-ambiguous orientation knowledge, and significantly improves the performance of lightweight detectors upon non-axially arranged objects. Extensive experiments on multiple datasets demonstrate that our plug-and-play distillation framework achieves state-of-the-art performance. Codes are available at https://github.com/Molly6/OD.