Hui Liu

Although deep learning methods have made significant strides in single image sand dust removal, the heterogeneous uncertainty induced by dusty environments poses a considerable challenge. In response, our research presents a novel framework known as the Hierarchical Interactive Uncertainty-aware Network (HIUNet). HIUNet leverages Bayesian neural networks for the extraction of robust shallow features, bolstered by pre-trained encoders for feature extraction and the agility of lightweight decoders for preliminary image reconstitution. Subsequently, a feature frequency selection mechanism is activated to enhance overall performance by strategically identifying and retaining valuable features while effectively suppressing redundant and irrelevant ones. Following this, a feature enhancement module is applied to the preliminary restoration. This intricate fusion culminates in the production of a restored image of superior quality. Our extensive experiments, using our proposed Sand11K dataset that exhibits various levels of degradation from dust and sand, confirm the effectiveness and soundness of our proposed method. By modeling uncertainty via Bayesian neural networks to extract robust shallow features and selecting valuable features through frequency selection, HIUNet can reconstruct high-quality clean images. For future work, we plan to extend our uncertainty-aware framework to handle extreme sand scenarios.

Within the domain of low-level vision, enhancing low-light images and removing sand-dust from single images are both critical tasks. These challenges are particularly pronounced in real-world applications such as autonomous driving, surveillance systems, and remote sensing, where adverse lighting and environmental conditions often degrade image quality. Various neural network models, including MLPs, CNNs, GANs, and Transformers, have been proposed to tackle these challenges, with the Vision KAN models showing particular promise. However, existing models, including the Vision KAN models use deterministic neural networks that do not address the uncertainties inherent in these processes. To overcome this, we introduce the Uncertainty-Aware Kolmogorov-Arnold Network (UAKAN), a novel structure that integrates KAN with uncertainty estimation. Our approach uniquely employs Tokenized KANs for sampling within a U-Net architecture’s encoder and decoder layers, enhancing the network’s ability to learn complex representations. Furthermore, for aleatoric uncertainty, we propose an uncertainty coupling certainty module that couples uncertainty distribution learning and residual learning in a feature fusion manner. For epistemic uncertainty, we propose a feature selection mechanism for spatial and pixel dimension uncertainty modeling, which captures and models uncertainty by learning the uncertainty contained between feature maps. Notably, our uncertainty-aware framework enables the model to produce both high-quality enhanced images and reliable uncertainty maps, which are crucial for downstream applications requiring confidence estimation. Through comparative and ablation studies on our synthetic SLLIE6K dataset, designed for low-light enhancement and sand-dust removal, we validate the effectiveness and theoretical robustness of our methodology.

In network real‐time kinematic positioning of multi‐reference station, the spatial and temporal distribution of tropospheric delay is affected by both horizontal and elevation. The traditional modeling strategy of regional troposphere takes more consideration of the horizontal factor, and the incomplete consideration of the elevation factor will lead to the problem of reduced modeling accuracy, especially in the face of the scene with large regional height deviation. Based on the traditional linear interpolation method (LIM), a simple and effective height‐deviation compensation‐based linear interpolation method (HCLIM) for regional tropospheric is proposed. The modeling accuracy of troposphere and the positioning accuracy of user RTK in large height deviation region are significantly improved. The method was verified based on six experimental subnets with large height deviations from a provincial continuously operating GNSS reference stations network in central China. The results showed that: For GPS satellite modeling, compared with the traditional LIM method, the average modeling accuracy improvement rate of HCLIM method is (84.5%, 75.5%, 59.3%, 26.7%) in the elevation angle range of (10–30°/30–40°/40–50°/50–90°). For BDS satellite, the average modeling accuracy improvement rate of HCLIM method in the above four elevation angles is (83.3%, 70%, 50%, 23.5%). For the positioning performance of user RTK, The horizontal positioning accuracy and RTK fixing rate were similar under the two methods, while HCLIM method showed only slight improvement. However, in the U direction, LIM method showed obvious systematic bias, while HCLIM method showed consistent positioning accuracy, which was improved to 82.8% compared with LIM method.

Introduction: Bacterial wilt (BW) caused by the aerobic, Gram-negative pathogenic species Ralstonia solanacearum (RS) is a major disease impacting commercial agriculture worldwide. Asian phylotype I of RS is the cause of tomato bacterial wilt, which has caused severe economic losses in southern China for many years. An urgent priority in control of bacterial wilt is development of rapid, sensitive, effective methods for detection of RS.Methods: We describe here a novel RS detection assay based on combination of loop-mediated isothermal amplification (LAMP) and CRISPR/Cas12a. crRNA1, with high trans-cleavage activity targeting hrpB gene, was selected out of four candidate crRNAs. Two visual detection techniques, involving naked-eye observation of fluorescence and lateral flow strips, were tested and displayed high sensitivity and strong specificity.Results and Discussion: The LAMP/Cas12a assay accurately detected RS phylotype Ⅰ in 14 test strains, and showed low detection limit (2.0 × 10⁰ copies). RS in tomato stem tissue and soil samples from two field sites with suspected BW infection was identified accurately, suggesting potential application of LAMP/Cas12a assay as point-of-care test (POCT). The overall detection process took less than 2 h and did not require professional lab equipment. Our findings, taken together, indicate that LAMP/Cas12a assay can be developed as an effective, inexpensive technique for field detection and monitoring of RS.

Extrinsic Calibration between LiDAR and POS (position and orientation system) is a fundamental prerequisite for varieties of MLS (mobile laser scanner) applications. Due to the sparse structure of LiDAR data, the current calibration methods relying on common point feature matching are unreliable, and the low accuracy POS results make the extrinsic calibration of MLS system more challenging. In this paper, we propose an incremental estimation method of six degree of freedom extrinsic transformation of LiDAR and POS. Firstly, the POS-SLAM is used to accumulate LiDAR scans as online sub maps. Attitudes of the carrier are calculated by using GNSS/INS loose combination method of bidirectional adjustment, and scans are associated with sub map based on the time interpolation. Then, the extrinsic calibration parameters are estimated by optimizing corresponding points difference between SLAM and MLS coordinate frame. Finally, field tests have been conducted to the proposed method. RMS between the map by the calibrated MLS and by the static measurement is 0.57 cm. The results demonstrate that the accuracy and robustness of our calibration approach are sufficient for mapping requirement of MLS.

High-precision and robust localization in GNSS-denied areas is crucial for autonomous vehicles and robots. Most state-of-the-art localization methods are based on simultaneous localization and mapping (SLAM) with a camera or light detection and ranging (LiDAR). However, SLAM will suffer from drift during long-term running without loop closure or prior constraints. Lightweight, high-precision environmental maps have gradually become an indispensable part of future autonomous driving. In order to solve the problem of real-time global localization for autonomous vehicles and robots, we propose a precise and robust LiDAR localization system based on a pre-built, occupied high-definition (HD) map called the Extended QuadTree (EQT) map. It makes use of a planar quadtree for block division and a Z-sequence index structure within the block cells. Then, a four-level occupancy probability cell value model is adopted. It will save about eight times the storage space compared with Google Cartographer, and the EQT map can be extended to store other information. For efficient scan-to-map matching with our specialized EQT map, the Bursa linearized model is used in the Gauss–Newton iteration of our algorithm, which makes the calculation of partial derivatives fast. All the above improvements lead to optimal storage and efficient querying for real-time scan-to-map matching localization. Field tests in an industrial park and road environment prove that positioning accuracy of about 6–13 cm and attitude accuracy of about 0.15° were achieved using a VLP-16 LiDAR. They also show that the method proposed in this paper is significantly better than the NDT method. For the long and narrow environment of an underground mine tunnel, high-resolution maps are also helpful for accurate and robust localization.

Many low Earth orbit (LEO) satellite constellations have been designed in recent years to provide global broadband Internet services. These constellations provide opportunities for LEO satellites to serve as navigation satellites by launching navigation signals while also being equipped with an onboard global navigation satellite system (GNSS) receiver. We propose a kinematic precise orbit determination (KPOD) and precise clock estimation (PCE) approach for LEO satellites by integrating regional ground observations and onboard observations of LEO satellites. By taking into account the LEO satellite clock bias, this approach can demonstrate the contributions of the ground and onboard observations of LEO satellites to the orbit and clock results, respectively. A composite LEO satellite constellation consisting of 168 satellites and observations from regional ground stations and LEO satellite onboard receivers are simulated considering the LEO satellite clock bias to verify the proposed approach. The results indicate that the convergence time of LEO satellite orbit determination can reach 9.38 min with the integrated KPOD (IKPOD) method, which is a reduction of 24.0% compared with the traditional KPOD method. However, the additional improvement in the LEO satellite orbit accuracy after convergence is very limited. The average root mean square (RMS) and standard deviation (STD) values of all LEO satellite clocks using the integrated PCE (IPCE) method with the participation of LEO satellite onboard observations can reach 0.27 ns and 0.15 ns, respectively. The improvements in the average RMS and STD are 42.6% and 60.5%, respectively, compared with the ground LEO PCE method. The convergence time and accuracy of LEO/GNSS precise point positioning can be improved by 44.6%, 48.3%, 26.7%, and 20.4% in the east, north, and up directions, respectively, using LEO satellite orbits and clocks from the IKPOD and IPCE methods compared with KPOD and PCE.