Search Results (39)

The complexity of land types and the limited spatial resolution of Synthetic Aperture Radar (SAR) imagery have led to widespread mixed-pixel contamination in radar backscatter images. The radar backscatter echo signals from a mixed pixel are often a combination of backscattering contributions from multiple endmembers. The signal mixture of endmembers within mixed pixels hinders the establishment of accurate relationships between pure endmembers’ parameters and the corresponding backscatter coefficient, thereby significantly reducing the accuracy of surface parameter inversion. However, few studies have focused on decomposing and estimating the pure backscatter signals within mixed pixels. This paper proposes a novel approach based on hyperspectral unmixing techniques and the microwave backscatter contribution decomposition (MBCD) model to estimate the pure backscatter coefficients of all Endmembers within mixed pixels. Experimental results demonstrate that the model performance varied significantly with endmember abundance. Specifically, high accuracy was achieved in estimating soil backscattering coefficients when vegetation coverage was below 25% ($R^2\approx 0.88$, with 98% of pixels showing relative errors within 0–20%); however, this accuracy declined as vegetation coverage increased. For grass endmembers, the model maintained high estimation precision across the entire grassland area (vegetation coverage 0.2–0.8), yielding an of 0.80 with 83% of pixels falling within the 0–20% relative error range. In addition, the model performance is influenced by the number of endmembers. Full article

The use of Uncrewed Aerial Systems (UASs) for remote sensing applications has increased significantly in recent years due to their low cost, operational flexibility, and rapid advancements in sensor technologies. In many cases, UAS platforms are considered viable alternatives to conventional satellite and crewed airborne platforms, offering very high spatial, spectral, and temporal resolution data. However, the radiometric quality of UAS-acquired data has not received equivalent attention, particularly with respect to absolute calibration. In this study, we (1) evaluate the absolute radiometric performance of two commonly used UAS sensors: the Headwall Nano-Hyperspec hyperspectral sensor and the MicaSense RedEdge-MX Dual Camera multispectral system; (2) assess the effectiveness of the Empirical Line Method (ELM) in improving the radiometric accuracy of reflectance products generated by these sensors; and (3) investigate the influence of calibration target characteristics—including size, material type, reflectance intensity, and quantity—on the performance of ELM for UAS data. A field campaign was conducted jointly by the U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) Center and the USGS National Uncrewed Systems Office (NUSO) from 15 to 18 July 2023, at the USGS EROS Ground Validation Radiometer (GVR) site in Sioux Falls, South Dakota, USA, over a 160 m × 160 m vegetated area. Absolute calibration accuracy was evaluated by comparing UAS sensor-derived reflectance to in situ measurements of the site. Results indicate that the Headwall Nano-Hyperspec and MicaSense sensors underestimated reflectance by approximately 0.05 and 0.015 reflectance units, respectively. While the MicaSense sensor demonstrated better inherent radiometric accuracy, it exhibited saturation over bright targets due to limitations in its automatic gain and exposure settings. Application of the ELM using just two calibration targets reduced discrepancies to within 0.005 reflectance units. Reflectance products generated using various target materials—such as felt, melamine, or commercially available validation targets—showed comparable agreement with in situ measurements when used with the Nano-Hyperspec sensor. Furthermore, increasing the number of calibration targets beyond two did not yield measurable improvements in calibration accuracy. At a flight altitude of 200 ft above ground level (AGL), a target size of 0.6 m × 0.6 m or larger was sufficient to provide pure pixels for ELM implementation, whereas smaller targets (e.g., 0.3 m × 0.3 m) posed challenges in isolating pure pixels. Overall, the standard manufacturer-recommended calibration procedures were insufficient for achieving high radiometric accuracy with the tested sensors, which may restrict their applicability in scenarios requiring greater accuracy and precision. The use of the ELM significantly improved data quality, enhancing the reliability and applicability of UAS-based remote sensing in contexts requiring high precision and accuracy. Full article

Hyperspectral unmixing aims to extract pure spectral signatures (endmembers) and estimate their corresponding abundance fractions from mixed pixels, enabling quantitative analysis of surface material composition. However, in geological mineral exploration, existing unmixing methods often fail to explicitly identify informative spectral bands, lack inter-layer information transfer mechanisms, and overlook the physical constraints intrinsic to the unmixing process. These issues result in limited directionality, sparsity, and interpretability. To address these limitations, this paper proposes a novel model, CResDAE, based on a deep autoencoder architecture. The encoder integrates a channel attention mechanism and deep residual modules to enhance its ability to assign adaptive weights to spectral bands in geological hyperspectral unmixing tasks. The model is evaluated by comparing its performance with traditional and deep learning-based unmixing methods on synthetic datasets, and through a comparative analysis with a nonlinear autoencoder on the Urban hyperspectral scene. Experimental results show that CResDAE consistently outperforms both conventional and deep learning counterparts. Finally, CResDAE is applied to GF-5 hyperspectral imagery from Yunnan Province, China, where it effectively distinguishes surface materials such as Forest, Grassland, Silicate, Carbonate, and Sulfate, offering reliable data support for geological surveys and mineral exploration in covered regions. Full article

Fractional Vegetation Cover (FVC) is a crucial indicator describing vegetation conditions and provides essential data for ecosystem health assessments. However, due to the low and sparse vegetation in alpine meadows, it is challenging to obtain pure vegetation pixels from Sentinel-2 imagery, resulting in errors in the FVC estimation using traditional pixel dichotomy models. This study integrated Sentinel-2 imagery with unmanned aerial vehicle (UAV) data and utilized the pixel dichotomy model together with four machine learning algorithms, namely Random Forest (RF), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Deep Neural Network (DNN), to estimate FVC in an alpine meadow region. First, FVC was preliminarily estimated using the pixel dichotomy model combined with nine vegetation indices applied to Sentinel-2 imagery. The performance of these estimates was evaluated against reference FVC values derived from centimeter-level UAV data. Subsequently, four machine learning models were employed for an accurate FVC inversion, using the estimated FVC values and UAV-derived reference FVC as inputs, following feature importance ranking and model parameter optimization. The results showed that: (1) Machine learning algorithms based on Sentinel-2 and UAV imagery effectively improved the accuracy of FVC estimation in alpine meadows. The DNN-based FVC estimation performed best, with a coefficient of determination of 0.82 and a root mean square error (RMSE) of 0.09. (2) In vegetation coverage estimation based on the pixel dichotomy model, different vegetation indices demonstrated varying performances across areas with different FVC levels. The GNDVI-based FVC achieved a higher accuracy (RMSE = 0.08) in high-vegetation coverage areas (FVC > 0.7), while the NIRv-based FVC and the SR-based FVC performed better (RMSE = 0.10) in low-vegetation coverage areas (FVC < 0.4). The method provided in this study can significantly enhance FVC estimation accuracy with limited fieldwork, contributing to alpine meadow monitoring on the Qinghai–Tibet Plateau. Full article

Non-photosynthetic vegetation (NPV) significantly impacts ecosystem degradation, drought, and wildfire risk due to its flammable and persistent litter. Yet, the accurate estimation of NPV in heterogeneous landscapes, such as dry–hot valleys, has been limited. This study utilized multi-source time-series remote sensing data from Sentinel-2 and GF-2, along with field surveys, to develop an NDVI-DFI ternary linear mixed model for quantifying NPV coverage (f_NPV) in a typical dry–hot valley region in 2023. The results indicated the following: (1) The NDVI-DFI ternary linear mixed model effectively estimates photosynthetic vegetation coverage (f_PV) and f_NPV, aligning well with the conceptual framework and meeting key assumptions, demonstrating its applicability and reliability. (2) The RGB color composite image derived using the minimum inclusion endmember feature method (MVE) exhibited darker tones, suggesting that MVE tends to overestimate the vegetation fraction when distinguishing vegetation types from bare soil. On the other hand, the pure pixel index (PPI) method showed higher accuracy in estimation due to its higher spectral purity and better recognition of endmembers, making it more suitable for studying dry–hot valley areas. (3) Estimates based on the NDVI-DFI ternary linear mixed model revealed significant seasonal shifts between PV and NPV, especially in valleys and lowlands. From the rainy to the dry season, the proportion of NPV increased from 23.37% to 35.52%, covering an additional 502.96 km². In summary, these findings underscore the substantial seasonal variations in f_PV and f_NPV, particularly in low-altitude regions along the valley, highlighting the dynamic nature of vegetation in dry–hot environments. Full article

In response to economic and environmental challenges like sea-level rise, salinity intrusion, groundwater extraction, sand mining, and sinking delta phenomena, the demand for solutions to adapt to changing conditions in riverine environments has increased significantly. High-quality analyses of land use and land cover (LULC) dynamics play a critical role in addressing these challenges. This study introduces a novel high-spatial resolution satellite-based approach to identify sub-seasonal LULC dynamics in the Mekong River Delta (MRD), employing a three-year (2021–2023) Sentinel-1 and Sentinel-2 satellite data time series. The primary obstacle is discerning detailed vegetation dynamics, particularly the seasonality of rice crops, answered through quantile mapping, harmonic regression with Fourier transform, and phenological metrics as inputs to a random forest machine learning classifier. Due to the substantial data volume, Google’s cloud computing platform Earth Engine was utilized for the analysis. Furthermore, the study evaluated the relative significance of various input features. The overall accuracy of the classification is 82.6% with a kappa statistic of 0.81, determined using comprehensive reference data collected in Vietnam. While the purely pixel-based approach has limitations, it proves to be a viable method for high-spatial resolution satellite image time series classification of the MRD. Full article

The accurate detection of electrical equipment states and faults is crucial for the reliable operation of such equipment and for maintaining the health of the overall power system. The state of power equipment can be effectively monitored through deep learning-based visual inspection methods, which provide essential information for diagnosing and predicting equipment failures. However, there are significant challenges: on the one hand, electrical equipment typically operates in complex environments, thus resulting in captured images that contain environmental noise, which significantly reduces the accuracy of state recognition based on visual perception. This, in turn, affects the comprehensiveness of the power system’s situational awareness. On the other hand, visual perception is limited to obtaining the appearance characteristics of the equipment. The lack of logical reasoning makes it difficult for purely visual analysis to conduct a deeper analysis and diagnosis of the complex equipment state. Therefore, to address these two issues, we first designed an image super-resolution reconstruction method based on the Generative Adversarial Network (GAN) to filter environmental noise. Then, the pixel information is analyzed using a deep learning-based method to obtain the spatial feature of the equipment. Finally, by constructing the logic diagram for electrical equipment clusters, we propose an interpretable fault diagnosis method that integrates the spatial features and temporal states of the electrical equipment. To verify the effectiveness of the proposed algorithm, extensive experiments are conducted on six datasets. The results demonstrate that the proposed method can achieve high accuracy in diagnosing electrical equipment faults. Full article

The estimation of depth in optically shallow waters using satellite imagery can be efficient and cost-effective. Active sensors measure the distance traveled by an emitted laser pulse propagating through the water with high precision and accuracy if the bottom peak intensity of the waveform is greater than the noise level. However, passive optical imaging of optically shallow water involves measuring the radiance after the sunlight undergoes downward attenuation on the way to the sea floor, and the reflected light is then attenuated while moving back upward to the water surface. The difficulty of satellite-derived bathymetry (SDB) arises from the fact that the measured radiance is a result of a complex association of physical elements, mainly the optical properties of the water, bottom reflectance, and depth. In this research, we attempt to apply physics-based algorithms to solve this complex problem as accurately as possible to overcome the limitation of having only a few known values from a multispectral sensor. Major analysis components are atmospheric correction, the estimation of water optical properties from optically deep water, and the optimization of bottom reflectance as well as the water depth. Specular reflection of the sky radiance from the water surface is modeled in addition to the typical atmospheric correction. The physical modeling of optically dominant components such as dissolved organic matter, phytoplankton, and suspended particulates allows the inversion of water attenuation coefficients from optically deep pixels. The atmospheric correction and water attenuation results are used in the ocean optical reflectance equation to solve for the bottom reflectance and water depth. At each stage of the solution, physics-based models and a physically valid, constrained Levenberg–Marquardt numerical optimization technique are used. The physics-based algorithm is applied to Landsat Operational Land Imager (OLI) imagery over the shallow coastal zone of Guam, Key West, and Puerto Rico. The SDB depths are compared to airborne lidar depths, and the root mean squared error (RMSE) is mostly less than 2 m over water as deep as 30 m. As the initial choice of bottom reflectance is critical, along with the bottom reflectance library, we describe a pure bottom unmixing method based on eigenvector analysis to estimate unknown site-specific bottom reflectance. Full article

Hyperspectral imaging captures detailed spectral data for remote sensing. However, due to the limited spatial resolution of hyperspectral sensors, each pixel of a hyperspectral image (HSI) may contain information from multiple materials. Although the hyperspectral unmixing (HU) process involves estimating endmembers, identifying pure spectral components, and estimating pixel abundances, existing algorithms mostly focus on just one or two tasks. Blind source separation (BSS) based on nonnegative matrix factorization (NMF) algorithms identify endmembers and their abundances at each pixel of HSI simultaneously. Although they perform well, the factorization results are unstable, require high computational costs, and are difficult to interpret from the original HSI. CUR matrix decomposition selects specific columns and rows from a dataset to represent it as a product of three small submatrices, resulting in interpretable low-rank factorization. In this paper, we propose a new blind HU framework based on CUR factorization called CUR-HU that performs the entire HU process by exploiting the low-rank structure of given HSIs. CUR-HU incorporates several techniques to perform the HU process with a performance comparable to state-of-the-art methods but with higher computational efficiency. We adopt a deterministic sampling method to select the most informative pixels and spectrum components in HSIs. We use an incremental QR decomposition method to reduce computation complexity and estimate the number of endmembers. Various experiments on synthetic and real HSIs are conducted to evaluate the performance of CUR-HU. CUR-HU performs comparably to state-of-the-art methods for estimating the number of endmembers and abundance maps, but it outperforms other methods for estimating the endmembers and the computational efficiency. It has a $9.4$ to $249.5$ times speedup over different methods for different real HSIs. Full article

With the development of artificial intelligence, the ability to capture the background characteristics of hyperspectral imagery (HSI) has improved, showing promising performance in hyperspectral anomaly detection (HAD) tasks. However, existing methods proposed in recent years still suffer from certain limitations: (1) Constraints are lacking in the deep feature learning process in terms of the issue of the absence of prior background and anomaly information. (2) Hyperspectral anomaly detectors with traditional self-supervised deep learning methods fail to ensure prioritized reconstruction of the background. (3) The architecture of fully connected deep networks in hyperspectral anomaly detectors leads to low utilization of spatial information and the destruction of the original spatial relationship in hyperspectral imagery and disregards the spectral correlation between adjacent pixels. (4) Hypotheses or assumptions for background and anomaly distributions restrict the performance of many hyperspectral anomaly detectors because the distributions of background land covers are usually complex and not assumable in real-world hyperspectral imagery. In consideration of the above problems, in this paper, we propose a novel fully convolutional auto-encoder based on dual clustering and latent feature adversarial consistency (FCAE-DCAC) for HAD, which is carried out with self-supervised learning-based processing. Firstly, density-based spatial clustering of applications with a noise algorithm and connected component analysis are utilized for successive spectral and spatial clustering to obtain more precise prior background and anomaly information, which facilitates the separation between background and anomaly samples during the training of our method. Subsequently, a novel fully convolutional auto-encoder (FCAE) integrated with a spatial–spectral joint attention (SSJA) mechanism is proposed to enhance the utilization of spatial information and augment feature expression. In addition, a latent feature adversarial consistency network with the ability to learn actual background distribution in hyperspectral imagery is proposed to achieve pure background reconstruction. Finally, a triplet loss is introduced to enhance the separability between background and anomaly, and the reconstruction residual serves as the anomaly detection result. We evaluate the proposed method based on seven groups of real-world hyperspectral datasets, and the experimental results confirm the effectiveness and superior performance of the proposed method versus nine state-of-the-art methods. Full article

Since camera and LiDAR sensors provide complementary information for the 3D semantic segmentation of intelligent vehicles, extensive efforts have been invested to fuse information from multi-modal data. Despite considerable advantages, fusion-based methods still have inevitable limitations: field-of-view disparity between two modal inputs, demanding precise paired data as inputs in both the training and inferring stages, and consuming more resources. These limitations pose significant obstacles to the practical application of fusion-based methods in real-world scenarios. Therefore, we propose a robust 3D semantic segmentation method based on multi-modal collaborative learning, aiming to enhance feature extraction and segmentation performance for point clouds. In practice, an attention based cross-modal knowledge distillation module is proposed to effectively acquire comprehensive information from multi-modal data and guide the pure point cloud network; then, a confidence-map-driven late fusion strategy is proposed to dynamically fuse the results of two modalities at the pixel-level to complement their advantages and further optimize segmentation results. The proposed method is evaluated on two public datasets (urban dataset SemanticKITTI and off-road dataset RELLIS-3D) and our unstructured test set. The experimental results demonstrate the competitiveness of state-of-the-art methods in diverse scenarios and a robustness to sensor faults. Full article

In recent years, health-monitoring systems have become increasingly important in the medical and safety fields, including patient and driver monitoring. Remote photoplethysmography is an approach that captures blood flow changes due to cardiac activity by utilizing a camera to measure transmitted or reflected light through the skin, but it has limitations in its sensitivity to changes in illumination and motion. Moreover, remote photoplethysmography signals measured from nonskin regions are unreliable, leading to inaccurate remote photoplethysmography estimation. In this study, we propose Skin-SegNet, a network that minimizes noise factors and improves pulse signal quality through precise skin segmentation. Skin-SegNet separates skin pixels and nonskin pixels, as well as accessories such as glasses and hair, through training on facial structural elements and skin textures. Additionally, Skin-SegNet reduces model parameters using an information blocking decoder and spatial squeeze module, achieving a fast inference time of 15 ms on an Intel i9 CPU. For verification, we evaluated Skin-SegNet using the PURE dataset, which consists of heart rate measurements from various environments. When compared to other skin segmentation methods with similar inference speeds, Skin-SegNet demonstrated a mean absolute percentage error of 1.18%, showing an improvement of approximately 60% compared to the 4.48% error rate of the other methods. The result even exhibits better performance, with only 0.019 million parameters, in comparison to DeepLabV3+, which has 5.22 million model parameters. Consequently, Skin-SegNet is expected to be employed as an effective preprocessing technique for facilitating efficient remote photoplethysmography on low-spec computing devices. Full article

Left ventricular ejection fraction (LVEF) plays as an essential role in the assessment of cardiac function, providing quantitative data support for the medical diagnosis of heart disease. Robust evaluation of the ejection fraction relies on accurate left ventricular (LV) segmentation of echocardiograms. Because human bias and expensive labor cost exist in manual echocardiographic analysis, computer algorithms of deep-learning have been developed to help human experts in segmentation tasks. Most of the previous work is based on the convolutional neural networks (CNN) structure and has achieved good results. However, the region occupied by the left ventricle is large for echocardiography. Therefore, the limited receptive field of CNN leaves much room for improvement in the effectiveness of LV segmentation. In recent years, Vision Transformer models have demonstrated their effectiveness and universality in traditional semantic segmentation tasks. Inspired by this, we propose two models that use two different pure Transformers as the basic framework for LV segmentation in echocardiography: one combines Swin Transformer and K-Net, and the other uses Segformer. We evaluate these two models on the EchoNet-Dynamic dataset of LV segmentation and compare the quantitative metrics with other models for LV segmentation. The experimental results show that the mean Dice similarity of the two models scores are 92.92% and 92.79%, respectively, which outperform most of the previous mainstream CNN models. In addition, we found that for some samples that were not easily segmented, whereas both our models successfully recognized the valve region and separated left ventricle and left atrium, the CNN model segmented them together as a single part. Therefore, it becomes possible for us to obtain accurate segmentation results through simple post-processing, by filtering out the parts with the largest circumference or pixel square. These promising results prove the effectiveness of the two models and reveal the potential of Transformer structure in echocardiographic segmentation. Full article

Spectral unmixing of geological mixtures, such as rocks, is a challenging inversion problem because of nonlinear interactions of light with the intimately mixed minerals at a microscopic scale. The fine-scale mixing of minerals in rocks limits the sensor’s ability to identify pure mineral endmembers and spectrally resolve these constituents within a given spatial resolution. In this study, we attempt to model the spectral unmixing of two rocks, namely, serpentinite and granite, by acquiring their hyperspectral images in a controlled environment, having uniform illumination, using a laboratory-based imaging spectroradiometer. The endmember spectra of each rock were identified by comparing a limited set of pure hyperspectral image pixels with the constituent minerals of the rocks based on their diagnostic spectral features. A series of spectral unmixing paradigms for explaining geological mixtures, including those ranging from simple physics-based light interaction models (linear, bilinear, and polynomial models) to classification-based models (support vector machines (SVMs) and half Siamese network (HSN)), were tested to estimate the fractional abundances of the endmembers at each pixel position of the image. The analysis of the results of the spectral unmixing algorithms using the ground truth abundance maps and actual mineralogical composition of the rock samples (estimated using X-ray diffraction (XRD) analysis) indicate a better performance of the pure pixel-guided HSN model in comparison to the linear, bilinear, polynomial, and SVM-based unmixing approaches. The HSN-based approach yielded reduced errors of abundance estimation, image reconstruction, and mineralogical composition for serpentinite and granite. With its ability to train using limited pure pixels, the half-Siamese network model has a scope for spectrally unmixing rock samples of varying mineralogical composition and grain sizes. Hence, HSN-based approaches effectively address the modelling of nonlinear mixing in geological mixtures. Full article

Developing complex hyperspectral image (HSI) sensors that capture high-resolution spatial information and voluminous (hundreds) spectral bands of the earth’s surface has made HSI pixel-wise classification a reality. The 3D-CNN has become the preferred HSI pixel-wise classification approach because of its ability to extract discriminative spectral and spatial information while maintaining data integrity. However, HSI datasets are characterized by high nonlinearity, voluminous spectral features, and limited training sample data. Therefore, developing deep HSI classification methods that purely utilize 3D-CNNs in their network structure often results in computationally expensive models prone to overfitting when the model depth increases. In this regard, this paper proposes an integrated deep multi-scale 3D/2D convolutional network block (MiCB) for simultaneous low-level spectral and high-level spatial feature extraction, which can optimally train on limited sample data. The strength of the proposed MiCB model solely lies in the innovative arrangement of convolution layers, giving the network the ability (i) to simultaneously convolve the low-level spectral with high-level spatial features; (ii) to use multiscale kernels to extract abundant contextual information; (iii) to apply residual connections to solve the degradation problem when the model depth increases beyond the threshold; and (iv) to utilize depthwise separable convolutions in its network structure to address the computational cost of the proposed MiCB model. We evaluate the efficacy of our proposed MiCB model using three publicly accessible HSI benchmarking datasets: Salinas Scene (SA), Indian Pines (IP), and the University of Pavia (UP). When trained on small amounts of training sample data, MiCB is better at classifying than the state-of-the-art methods used for comparison. For instance, the MiCB achieves a high overall classification accuracy of 97.35%, 98.29%, and 99.20% when trained on 5% IP, 1% UP, and 1% SA data, respectively. Full article