Landslides, the second largest geological hazard after earthquakes, result in significant loss of life and property. Extracting landslide information quickly and accurately is the basis of landslide disaster prevention. Fengjie County, Chongqing, China, is a typical landslide-prone area in the Three Gorges Reservoir Area. In this study, we newly integrate Shapley Additive Explanation (SHAP) and Optuna (OPT) hyperparameter tuning into four basic machine learning algorithms: Gradient Boosting Decision Tree (GBDT), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Additive Boosting (AdaBoost). We construct four new models (SHAP-OPT-GBDT, SHAP-OPT-XGBoost, SHAP-OPT-LightGBM, and SHAP-OPT-AdaBoost) and apply the four new models to landslide extraction for the first time. Firstly, high-resolution remote sensing images were preprocessed, landslide and non-landslide samples were constructed, and an initial feature set with 48 features was built. Secondly, SHAP was used to select features with significant contributions, and the important features were selected. Finally, Optuna, the Bayesian optimization technique, was utilized to automatically select the basic models’ best hyperparameters. The experimental results show that the accuracy (ACC) of these four SHAP-OPT models was above 92% and the training time was less than 1.3 s using mediocre computational hardware. Furthermore, SHAP-OPT-XGBoost achieved the highest accuracy (96.26%). Landslide distribution information in Fengjie County from 2013 to 2020 can be extracted by SHAP-OPT-XGBoost accurately and quickly.

Accurate and high-spatiotemporal-resolution predictions of precipitable water vapor (PWV) play a crucial role in numerous atmospheric processes, including global navigation satellite system (GNSS) meteorological evaluation, weather forecasting, water cycle analysis, climate trend detection, and related fields. The calculation of PWV typically relies on meteorological parameters and conversion coefficients, leading to low accuracy and dependence on the availability of meteorological data. This research has formulated a high-accuracy, spatiotemporal-resolution PWV retrieval model that leverages non-meteorological parameters through the implementation of a multi-layer perceptron (MLP) neural network. The new retrieval model utilizes spatiotemporal information and the zenith tropospheric delay (ZTD) datasets from each GNSS station as training input parameters, with numerical weather model PWV serving as the training output parameters. The MLP neural network was trained using datasets from 229 stations of the Crustal Movement Observation Network of China (CMONOC) and numerical weather model PWV data collected from 2005 to 2015. The experimental results indicate that the root mean square (RMS) fitting accuracy between MLP PWV and ERA PWV is 1.36 mm. The MLP PWV model demonstrates an accuracy of approximately 1.6 mm on both temporal and spatial scales, with a spatiotemporal RMS accuracy indicator of 1.92 mm. The testing accuracy of the MLP PWV, approximately 2 mm when compared with the numerical meteorological parameters model PWV, underscores the reliability of this work. The MLP PWV retrieval model, relying on non-meteorological parameters, has demonstrated exceptional accuracy. By incorporating high spatiotemporal zenith tropospheric delay (ZTD), station spatial data, and temporal information, this model achieves superior spatiotemporal PWV accuracy, showcasing the effectiveness of the methodology presented in this study.

In unmanned aerial vehicle (UAV)-based monocular inspection, cracks typically present as geometrically asymmetric, elongated, low-contrast weak targets, making accurate segmentation and spatial localization challenging. Existing methods are susceptible to missed detections and false positives when handling slender cracks, and monocular 3D reconstruction for localization is often burdened by redundant frames, resulting in limited modeling efficiency. To mitigate these issues, we propose a high-precision framework for crack segmentation and spatial localization from UAV imagery. First, Oriented FAST and Rotated BRIEF–Simultaneous Localization and Mapping, version 3 (ORB-SLAM3) is adopted for keyframe selection to suppress data redundancy and improve reconstruction stability. Second, we develop an enhanced YOLOv11-seg model by integrating the Dilation-wise Residual Segmentation (DWRSeg) module, the Weighted IoU (WIoU) loss, and the Lightweight shared convolutional separator batch-normalization detection head (LSCSBD) to strengthen feature discrimination and segmentation robustness for slender cracks, yielding high-quality crack masks. Finally, the predicted masks are projected onto the reconstructed 3D surface to obtain precise spatial localization. Our experimental results demonstrate that the proposed approach improves the segmentation mAP@50 by 7.2% over the baseline while reducing computational complexity from 10.2 to 9.8 GFLOPs. In addition, keyframe-based processing reduces the 3D modeling time by 59.4% compared to that with full-frame reconstruction. Overall, the proposed framework jointly enhances crack segmentation accuracy and substantially accelerates 3D modeling and localization, providing an effective solution for efficient UAV-based crack inspection.

Being influenced by fluctuating precipitation, long-term rainfall-induced landslide displacement prediction could be unstable. To solve this problem, a rainfall-induced landslide displacement prediction model based on Attention Mechanism Neural Network(AMNN) is proposed in this paper. Firstly, accumulative landslide displacement is decomposed into the trend term and the periodic term. Secondly, multivariate linear regression is adopted to fit the trend term and AMNN is used to predict the periodic term. Finally, the accumulative predicted displacement is given by summing the trend and the periodic displacement components. In this paper, one rainfall-induced landslide in Chongqing province, China was taken as an example to evaluate the designed model. Compared with some existing methods, the model proposed in this paper can captured the correlation between each feature sequence and the predicted term with higher prediction accuracy of 0.97 Goodness of Fit. The results of this paper are believed to be contributive to further rainfall-induced landslide forecast and early warning research.

Small rotorcraft unmanned aerial vehicles (UAVs) are highly susceptible to wind disturbances when performing tasks such as fixed-point hovering, low-altitude inspection, and aggressive maneuvers. Under complex, variable meteorological conditions, attitude stability and position-holding accuracy are particularly critical. Although quadrotor UAVs exhibit structural and dynamic symmetry, real wind disturbances are often asymmetric, disrupting the original balance and leading to intensified attitude oscillations, position drift, and degraded data quality. To effectively address the challenges of wind-induced oscillation and positional deviation, this paper proposes a fuzzy logic-based linear active disturbance rejection control (Fuzzy-LADRC) strategy. This approach employs a hybrid algorithm combining particle swarm optimization and gray wolf optimization to optimize controller parameters and incorporates fuzzy logic to enhance the adaptive capability of the linear active disturbance rejection controller (LADRC). Simulation experiments conducted in MATLAB/Simulink under complex wind-field conditions demonstrate that the proposed method significantly outperforms traditional PID controllers: in the regulation of roll and pitch angles, control performance improves by approximately 5%, while in yaw angle control, the improvement reaches up to 30%. Furthermore, this method can significantly suppress position deviation and fluctuation in the X and Y directions, and reduce the overshoot in the Z-axis during the UAV’s takeoff phase by 75%.