Cracks are one of the main distresses that occur on concrete surfaces. Traditional methods for detecting cracks based on two-dimensional (2D) images can be hampered by stains, shadows, and other artifacts, while various three-dimensional (3D) crack-detection techniques, using point clouds, are less affected in this regard but are limited by the measurement accuracy of the 3D laser scanner. In this study, we propose an automatic crack-detection method that fuses 3D point clouds and 2D images based on an improved Otsu algorithm, which consists of the following four major procedures. First, a high-precision registration of a depth image projected from 3D point clouds and 2D images is performed. Second, pixel-level image fusion is performed, which fuses the depth and gray information. Third, a rough crack image is obtained from the fusion image using the improved Otsu method. Finally, the connected domain labeling and morphological methods are used to finely extract the cracks. Experimentally, the proposed method was tested at multiple scales and with various types of concrete crack. The results demonstrate that the proposed method can achieve an average precision of 89.0%, recall of 84.8%, and F1 score of 86.7%, performing significantly better than the single image (average F1 score of 67.6%) and single point cloud (average F1 score of 76.0%) methods. Accordingly, the proposed method has high detection accuracy and universality, indicating its wide potential application as an automatic method for concrete-crack detection.

Rail fasteners are the most numerous components in railways and they should be inspected periodically. Manual inspection is currently a common solution, which is laborious and low-efficient. Some automatic inspection approaches are proposed. But for ballast railway fasteners inspection, debris, especially ballast along tracks may cover the fasteners, which is still a tricky problem. In this paper, a real-time inspection system for ballast railway fasteners based on point cloud deep learning is developed. Dense and precise point cloud of fastener is obtained from the structured light sensors in the system. The point cloud of fastener is segmented into different parts to avoid the interference of debris on fasteners. A ballast fastener point cloud semantic segmentation dataset is created based on automatic annotation method. Several deep learning point cloud segmentation models are tested in this dataset and PointNet++ is selected to be deployed in the real-time deep learning module of the system. Field tests on ballast railways show excellent accuracy and efficiency of this system.

In the application of quantitative remote sensing in water quality monitoring, the existence of mixed pixels greatly affects the accuracy of water quality parameter inversion, especially for narrow inland rivers. Improving the image spatial resolution and weakening the interference of mixed pixels in the image are some of the urgent problems to be solved in the study of water quality monitoring of medium- and small-sized inland rivers. We processed Sentinel-2 multispectral images using the super-resolution algorithm and generated a set of 10 m spatial resolution images with basically unchanged reflection characteristics. Both qualitative and quantitative evaluation results show that the super-resolution algorithm can weaken the influence of mixed pixels while maintaining spectral invariance. Before the application of the super-resolution algorithm, the inversion accuracy of water quality parameters in this study were as follows: for NH₃-N, the R² was 0.61, the root mean squared error (RMSE) was 0.177 and the mean absolute percentage error (MAPE) was 29.33%; for Chemical Oxygen Demand (COD), the R² was 0.26, the RMSE was 0.756 and the MAPE was 4.62%; for Total Phosphorus (TP), the R² was 0.69, the RMSE was 0.032 and the MAPE was 30.58%. After the application of the super-resolution algorithm, the inversion accuracy of water quality parameters in this study were as follows: for NH₃-N, the R² was 0.67, the RMSE was 0.161 and the MAPE was 25.88%; for COD, the R² was 0.53, the RMSE was 0.546 and the MAPE was 3.36%; for TP, the R² was 0.60, the RMSE was 0.034 and the MAPE was 24.28%. Finally, the spatial distribution of NH₃-N, COD and TP was obtained by using a machine learning model. The results showed that the application of the super-resolution algorithm can effectively improve the retrieval accuracy of NH₃-N, COD and TP, which illustrates the application potential of the super-resolution algorithm in water quality remote sensing quantitative monitoring.

The dislocation of ring segments in shield tunnels poses adverse effects on tunnel structural stability and waterproofing. Current methods can only detect limited types of dislocation from circular tunnels, posing challenges in meeting practical requirements. This article introduces a dislocation detection method that is applicable to shield tunnels of various shapes, enabling the detection of both inter-ring and intra-ring dislocations. First, a point cloud lossless unfolding method is introduced, allowing for the unfolding of tunnel point clouds of any cross-sectional shape without compromising point cloud features. Subsequently, a subway tunnel point cloud segmentation workflow is proposed, enabling the direct application of image deep learning (DL) networks to tunnel point clouds. The bolts are then identified as markers for dislocation measurement, and a method for bolt area of interest (AOI) positioning and pairing is introduced. This method calculates dislocation value through fitting reference surfaces from AOI. A series of experiments conducted on circular and quasi-rectangular shield tunnels totaling 2.45 km and containing 20 billion points demonstrate that the proposed segmentation workflow, coupled with an image semantic segmentation model, achieves a maximum mean intersection over union (mIoU) of 90.2%, which is 10% higher than directly segmenting tunnel point clouds. The proposed method went through repeated accuracy tests and was compared with other methods, including total station measurement, RANSAC plane fitting, and scanline method. The results of our proposed method demonstrate high precision and stability, with a measurement accuracy RMSE of 0.89 mm, an MAE of 0.69 mm, and an efficiency of 20 times higher compared to total station measurement.

A point clouds filtering algorithm is presented based on Grid Partition using Dynamic Quad Tress and Moving Least Squares. First, points are partitioned reasonably and corresponding Dynamic Quad Trees indices are established. Second, points in grids are utilized to fit a DEM reference plane using moving least squares technology. Finally, ground points are separated from those non-ground ones if they are positioned above the reference plane and have a distance to the plane exceeding threshold value. Experiments show that this filtering algorithm is of high precision and identify ground points effectively without losing detailed topography information

Efficient semantic segmentation of large-scale 3D point clouds is a fundamental and essential capability for real-time intelligent systems, such as autonomous driving and augmented reality. The high dimension feature vector and the complex network structure are two major constraints to utilize the large-scale point cloud. This paper proposes an optimal feature description vector network (OFDV-Net) for 3D point cloud semantic segmentation. First, a multiscale point cloud feature extraction structure is constructed to generate an initial feature description vector (IFDV). Then, IFDV is selected by a feature selection unit to obtain the optimal feature description vector (OFDV). The OFDV encapsulates the best 3D features set of the points and can be used as the input of the deep neural network for training and testing. Finally, the OFDV-Net was applied to the standard public outdoor large-scale point cloud datasets Semantic3D and NPM3D, and the overall segmentation accuracy of 88.3% and 87.7% were obtained, respectively; moreover, the OFDV-Net requires less training time, which indicates that the algorithm can obtain high-precision semantic segmentation results on an outdoor large-scale point cloud while reducing model training time.