Search

High-precision and rapid detection of apparent defects in subway tunnel linings is crucial for ensuring the structural integrity of tunnels and the safety of train operations. However, current methods often do not adequately account for the spatial characteristics of these defects and perform poorly in detecting and extracting small-scale defects, which limits the accuracy of detection and geometric parameter extraction. To address these challenges, this paper proposes an efficient algorithm for detecting and extracting apparent defects in subway tunnels. Firstly, YOLOv8 was selected as the foundational architecture due to its comprehensive performance. The coordinate attention module and Bottleneck Transformer 3 were then integrated into the model’s backbone to enhance the focus on defect-prone areas and improve the learning of feature relationships between defects and other infrastructure. Subsequently, a high-resolution detection layer was added to the model’s head to further improve sensitivity to subtle defects. Additionally, a low-quality crack dataset was created using an open access dataset, and transfer learning combined with Real-ESRGAN was employed to enhance the detail and resolution of fine cracks. The results of the field experiments demonstrate that the proposed model significantly improves detection accuracy in high-incidence areas and for small-scale defects, achieving a mean average precision (mAP) of 87% in detecting cracks, leakage, exfoliation, and related infrastructure defects. Furthermore, the crack enhancement techniques substantially improve the representation of fine-crack details, increasing feature extraction accuracy by a factor of four. The findings of this paper could provide crucial technical support for the automated operation and maintenance of metro tunnels.

The failures of culverts can result in water and soil loss in the surrounding subsurface, leading to collapses. In densely constructed urban areas, buried culverts are inaccessible and invisible, making Ground Penetrating Radar (GPR) inspection a widely adopted nondestructive and efficient method. Previous research has employed artificial intelligence to automate GPR interpretation by identifying special radargram textures, but the irregular GPR responses of degraded subsurface soils present a challenge. Therefore, this study proposes a method for preliminary diagnosing large-scale culverts based on GPR wave attributes. The method comprises three parts: (1) evaluating soil compactness by estimating the tortuosity of GPR responses at the culvert top; (2) investigating soil moisture using the attenuation rate of GPR responses; and (3) overlaying the results of the incompact and moist soils to pinpoint potential degradations. Two long-distance culverts were inspected and diagnosed using the proposed method, and three abnormal sections were identified and confirmed by excavation evaluation. The case studies indicate that the proposed method can quickly provide reference for further in-depth investigation, thereby improving GPR survey efficiency and reducing workload by narrowing the investigation scope. Consequently, this study can facilitate large-scale GPR culvert surveys and further safeguard the water system.

As one of the most densely populated and rapidly growing metropolitan areas worldwide, the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) of China has been developing extensive metro networks to relieve the escalating traffic congestion inner cities and to shorten public transport time inter cities. Metro construction possibly triggers ground deformation, which may result in damage to the tunnel, pipeline, and ground structures. Fast and efficient monitoring ground deformation along metro lines is vital not only to the metro itself but also to these structures. This study provides a comprehensive investigation of ground deformation along 67 metro lines within the GBA, based on multi-temporal synthetic aperture radar interferometry (MT-InSAR) by using Sentinel-1 from March 2017 to February 2022. The results reveal that deformation velocity in this area ranges from –39.4 mm/year to 14.2 mm/year, which is mainly caused by the tunnel excavation, adjacent excavation and construction, and unfavorable geological conditions. Furthermore, risk levels have been evaluated based on the InSAR-derived results with the proposed median absolute deviation heatmap clustering method. The risky area accounts for 16.31% (∼205.24 km2) of the whole study area in the GBA, among which 13.76%, 2.16%, and 0.39% are classified as three risk levels, with mean deformation velocities of –2.2 mm/year, –3.4 mm/year, and –6.5 mm/year, respectively. This study presents a comprehensive megalopolitan-scale ground deformation monitoring, for the first time, along metro lines in six cities of the GBA, which provides important insights into future metro line constructing, operating and planning.

The surface defects on a shield subway tunnel can significantly affect the serviceability of the tunnel structure and may compromise operation safety. To effectively detect multiple surface defects, this study uses a tunnel inspection trolley (TIT) based on the mobile laser scanning technique. By conducting an inspection of the shield tunnel on a metro line section, various surface defects are identified with the TIT, including water leakage defects, dislocation, spalling, cross-section deformation, etc. To explore the root causes of the surface defects, association rules between different defects are calculated using an improved Apriori algorithm. The results show that: (i) there are significant differences in different association rules for various surface defects on the shield tunnel; (ii) the average confidence of the association rule “dislocation & spalling → water leakage” is as high as 57.78%, indicating that most of the water leakage defects are caused by dislocation and spalling of the shield tunnel in the sections being inspected; (iii) the weakest rule appears at “water leakage → spalling”, with an average confidence of 13%. The association analysis can be used for predicting the critical defects influencing structural reliability and operation safety, such as water leakage, and optimizing the construction and maintenance work for a shield subway tunnel.

Most existing sewer pipeline condition assessment methods determine the presence and types of faults via examination of videos, which is a time-consuming and labor-intensive process. A few automatic methods based on image processing techniques can be used to detect specific faults. However, these methods have limitations due to the presence of unpredictable sewer pipeline fault patterns. Deep learning methods have also been applied to sewer pipeline fault detection. However, these methods require a large amount of annotated data to obtain reliable results. In this paper, we propose a fault detection method that applies unsupervised machine learning based anomaly detection algorithms with feature extraction to videos recorded by new sewer pipeline visual inspection equipment. The recorded videos are regarded as sequence signals, which are converted into feature vectors, followed by application of an anomaly detection algorithm. Unlike existing methods, the proposed method is computationally efficient as it does not require an annotated fault sample database for training fault detection models. We evaluate various anomaly detection algorithms and feature combinations on real sewer pipeline data collected in Shenzhen, with an overall accuracy result of above 90%. The proposed method provides a new and fast technique for surveying urban sewer pipelines, and to facilitate further research in this area, we have made the code and data used in this paper publicly available.

We derived the joint slip models of the three major events in the 2016 Central Italy earthquake sequence by inverting strong-motion and InSAR datasets. b-values and the historic earthquake scarp offset were also investigated after processing the earthquake catalog and near-field digital elevation model data. The three major events gradually released seismic moments of 1.6 × 1018 Nm (Mw 6.1), 1.5 × 1018 Nm (Mw 6.1), and 1.1 × 1019 Nm (Mw 6.7), respectively. All the ruptures exhibit both updip and along-strike directivity, but differ in the along-strike propagation direction. The high b-value found beneath three mainshock hypocenters suggests possible fluid intrusions, explaining the cascading earthquake behavior. The cumulative surface scarp from past earthquakes shows rupturing features that are consistent with the 2016 earthquake sequence, suggesting a characteristic fault behavior. Under the assumption of the Gutenberg–Richter law, the slip budget closure test gives a maximum magnitude of Mw 6.7 and implies the seismic hazard from the largest event has been released in this sequence.

Visual odometry is the task of estimating the trajectory of the moving agents from consecutive images. It is a hot research topic both in robotic and computer vision communities and facilitates many applications, such as autonomous driving and virtual reality. The conventional odometry methods predict the trajectory by utilizing the multiple view geometry between consecutive overlapping images. However, these methods need to be carefully designed and fine-tuned to work well in different environments. Deep learning has been explored to alleviate the challenge by directly predicting the relative pose from the paired images. Deep learning-based methods usually focus on the consecutive images that are feasible to propagate the error over time. In this paper, graph loss and geodesic rotation loss are proposed to enhance deep learning-based visual odometry methods based on graph constraints and geodesic distance, respectively. The graph loss not only considers the relative pose loss of consecutive images, but also the relative pose of non-consecutive images. The relative pose of non-consecutive images is not directly predicted but computed from the relative pose of consecutive ones. The geodesic rotation loss is constructed by the geodesic distance and the model regresses a Lie algebra so(3) (3D vector). This allows a robust and stable convergence. To increase the efficiency, a random strategy is adopted to select the edges of the graph instead of using all of the edges. This strategy provides additional regularization for training the networks. Extensive experiments are conducted on visual odometry benchmarks, and the obtained results demonstrate that the proposed method has comparable performance to other supervised learning-based methods, as well as monocular camera-based methods. The source code and the weight are made publicly available.

Urban road intersections are one of the key components of road networks. Due to complex and diverse traffic conditions, traffic conflicts occur frequently. Accurate traffic conflict detection allows improvement of the traffic conditions and decreases the probability of traffic accidents. Many time-based conflict indicators have been widely studied, but the sizes of the vehicles are ignored. This is a very important factor for conflict detection at urban intersections. Therefore, in this paper we propose a novel time difference conflict indicator by incorporating vehicle sizes instead of viewing vehicles as particles. Specially, we designed an automatic conflict recognition framework between vehicles at the urban intersections. The vehicle sizes are automatically extracted with the sparse recurrent convolutional neural network, and the vehicle trajectories are obtained with a fast-tracking algorithm based on the intersection-to-union ratio. Given tracking vehicles, we improved the time difference to the conflict metric by incorporating vehicle size information. We have conducted extensive experiments and demonstrated that the proposed framework can effectively recognize vehicle conflict accurately.

Ground deformation related to mining activities may occur immediately or many years later, leading to a series of mine geological disasters, such as ground fissures, collapses, and even mining earthquakes. Deformation monitoring has been carried out with techniques, such as multitemporal interferometric synthetic aperture radar (MTInSAR). Over the past decade, MTInSAR has been widely used in monitoring mining deformation, and it is still difficult to retrieve mining deformation over dense vegetation areas. In this study, we use multiple-platform SAR images to retrieve mining deformation over dense vegetation areas. The high-quality interferograms are selected by the coherence map, and the mining deformation is retrieved by the MSBAS-InSAR technique. SAR images from TerraSAR-X, Sentinel-1A, Radarsat-2, and PALSAR-2 over the Fengfeng mining area, Heibei, China, are used to retrieve the deformation of mining activities covered with dense vegetation. The results show that the subsidence in the Fengfeng mining area reaches up to 90 cm over the period from July 2015 to April 2016. The root-mean-square error (RMSE) between the results from InSAR and leveling is 83.5 mm/yr at two mining sites, i.e., Wannian and Jiulong Mines.

Fisheye cameras are widely used in visual localization due to the advantage of the wide field of view. However, the severe distortion in fisheye images lead to feature matching difficulties. This paper proposes an IMU-assisted fisheye image matching method called spherically optimized random sample consensus (So-RANSAC). We converted the putative correspondences into fisheye spherical coordinates and then used an inertial measurement unit (IMU) to provide relative rotation angles to assist fisheye image epipolar constraints and improve the accuracy of pose estimation and mismatch removal. To verify the performance of So-RANSAC, experiments were performed on fisheye images of urban drainage pipes and public data sets. The experimental results showed that So-RANSAC can effectively improve the mismatch removal accuracy, and its performance was superior to the commonly used fisheye image matching methods in various experimental scenarios.

Ocean waves are a vital environmental factor that affects the accuracy of airborne laser bathymetry (ALB) systems. As the regional water surface undulates with randomness, the laser propagation direction through the air–water surface will change and impact the underwater topographic result from the ALB system, especially for the small laser divergence system. However, the natural ocean surface changes rapidly over time, and uneven ocean surface point clouds from ALB scanning will cause an uncertain estimation of the laser propagation direction; therefore, a self-adaptive correction method based on the characteristics of the partial wave surface is key to improving the accuracy and applicability of the ALB system. In this paper, we focused on the issues of spatial position deviation caused by surface waves and position correction of the underwater laser footprint, and the dimension-based adaptive method is applied to attempt to correct the laser incidence angle. Simulation experiments and analysis of the actual measurement data from different ALB systems verified that the method can effectively suppress the influence of ocean waves. Furthermore, the inversion result of sea surface inclination changes is consistent with the surface wind wave reanalysis products. Based on the laser underwater propagation model in the strategy, we also quantitatively analyzed the influence of surface waves on laser bathymetry, which can guide the operation selection and data processing of the ALB system at specific water depths and under dynamic ocean conditions.

For a city to be livable and walkable is the ultimate goal of future cities. However, conflicts among pedestrians, vehicles, and cyclists at traffic intersections are becoming severe in high-density urban transportation areas, especially in China. Correspondingly, the transit time at intersections is becoming prolonged, and pedestrian safety is becoming endangered. Simulating pedestrian movements at complex traffic intersections is necessary to optimize the traffic organization. We propose an unmanned aerial vehicle (UAV)-based method for tracking and simulating pedestrian movements at intersections. Specifically, high-resolution videos acquired by a UAV are used to recognize and position moving targets, including pedestrians, cyclists, and vehicles, using the convolutional neural network. An improved social force-based motion model is proposed, considering the conflicts among pedestrians, cyclists, and vehicles. In addition, maximum likelihood estimation is performed to calibrate an improved social force model. UAV videos of intersections in Shenzhen are analyzed to demonstrate the performance of the presented approach. The results demonstrate that the proposed social force-based motion model can effectively simulate the movement of pedestrians and cyclists at road intersections. The presented approach provides an alternative method to track and simulate pedestrian movements, thus benefitting the organization of pedestrian flow and traffic signals controlling the intersections.

This paper presents a novel indoor topological localization method based on mobile phone videos. Conventional methods suffer from indoor dynamic environmental changes and scene ambiguity. The proposed Visual Landmark Sequence-based Indoor Localization (VLSIL) method is capable of addressing problems by taking steady indoor objects as landmarks. Unlike many feature or appearance matching-based localization methods, our method utilizes highly abstracted landmark sematic information to represent locations and thus is invariant to illumination changes, temporal variations, and occlusions. We match consistently detected landmarks against the topological map based on the occurrence order in the videos. The proposed approach contains two components: a convolutional neural network (CNN)-based landmark detector and a topological matching algorithm. The proposed detector is capable of reliably and accurately detecting landmarks. The other part is the matching algorithm built on the second order hidden Markov model and it can successfully handle the environmental ambiguity by fusing sematic and connectivity information of landmarks. To evaluate the method, we conduct extensive experiments on the real world dataset collected in two indoor environments, and the results show that our deep neural network-based indoor landmark detector accurately detects all landmarks and is expected to be utilized in similar environments without retraining and that VLSIL can effectively localize indoor landmarks.

Urban land use is key to rational urban planning and management. Traditional land use classification methods rely heavily on domain experts, which is both expensive and inefficient. In this paper, deep neural network-based approaches are presented to label urban land use at pixel level using high-resolution aerial images and ground-level street view images. We use a deep neural network to extract semantic features from sparsely distributed street view images and interpolate them in the spatial domain to match the spatial resolution of the aerial images, which are then fused together through a deep neural network for classifying land use categories. Our methods are tested on a large publicly available aerial and street view images dataset of New York City, and the results show that using aerial images alone can achieve relatively high classification accuracy, the ground-level street view images contain useful information for urban land use classification, and fusing street image features with aerial images can improve classification accuracy. Moreover, we present experimental studies to show that street view images add more values when the resolutions of the aerial images are lower, and we also present case studies to illustrate how street view images provide useful auxiliary information to aerial images to boost performances.

Use of a frequency-division multiple access strategy causes Globalnaya Navigatsionnaya Sputnikovaya Sistema (GLONASS) receiving equipment to experience both inter-frequency phase bias (IFPB) and inter-frequency code bias (IFCB). While IFPB can be calibrated using a linear model, there is no general model for IFCB calibration, which causes great difficulty in GLONASS ambiguity resolution over long baselines; most current GLONASS ambiguity resolution research is confined to short baselines. In this paper, based on a single-differencing between-receivers (SDBR) model, a wide-lane phase combination-based approach is proposed to fix the GLONASS ambiguities over long baselines. External precise ionospheric products are introduced to eliminate the ionospheric delay. To mitigate the effect of the residual ionospheric delays, we fix the relative wide-lane ambiguity using the Hatch–Melbourne–Wubbena (HMW) combination. The results show that 96% and 55% of the wide-lane round-off residuals are within 0.2 cycles for the Global Positioning System (GPS) and GLONASS, respectively, if the traditional HMW method is used. The method proposed here for GLONASS can improve these percentages significantly, reaching up to 95.5%. The root mean square (RMS) position errors are 1.43, 1.06 and 4.32 mm for GPS in the north, east and up directions, respectively. When GLONASS with ambiguity fixing is added, the corresponding RMS values are reduced significantly to 1.26, 1.02 and 3.87 mm, respectively.

Vehicle behavior recognition is an attractive research field which is useful for many computer vision and intelligent traffic analysis tasks. This paper presents an all-in-one behavior recognition framework for moving vehicles based on the latest deep learning techniques. Unlike traditional traffic analysis methods which rely on low-resolution videos captured by road cameras, we capture 4K ( 3840 × 2178 ) traffic videos at a busy road intersection of a modern megacity by flying a unmanned aerial vehicle (UAV) during the rush hours. We then manually annotate locations and types of road vehicles. The proposed method consists of the following three steps: (1) vehicle detection and type recognition based on deep neural networks; (2) vehicle tracking by data association and vehicle trajectory modeling; (3) vehicle behavior recognition by nearest neighbor search and by bidirectional long short-term memory network, respectively. This paper also presents experimental results of the proposed framework in comparison with state-of-the-art approaches on the 4K testing traffic video, which demonstrated the effectiveness and superiority of the proposed method.

Correction of the fractional cycle bias (FCB) in the undifferenced ambiguity allows precise point positioning (PPP) integer ambiguity resolution (IAR) to be achieved, which can improve positioning accuracy significantly. In addition, in real-time PPP-IAR, integration of the BeiDou Navigation Satellite System (BDS) can provide a significant reduction in the initial fixing time of global positioning system (GPS)-only PPP-IAR. However, the FCB quality can be considerably affected by the low precision of the BDS orbit, which then severely hampers the GPS + BDS PPP-IAR performance. Therefore, a real-time FCB estimation strategy that takes the BDS satellite orbit error into consideration was developed in this study. The slant orbit error can be absorbed by the ionosphere-free (IF) ambiguity, which can then be recovered by fixing all IF ambiguities from all the tracking stations. The estimated orbit error is then used to refine the orbit, which is broadcast along with the FCBs to enable PPP ambiguity resolution. To evaluate the proposed strategy, an experiment using 60 tracking stations covering the China region is performed in a simulated real-time mode. The a posteriori residuals of both the wide- and narrow-lane ambiguities are checked to validate the efficiency of the proposed FCB strategy. The results show that when the proposed strategy is applied, the effect of the BDS orbit error on narrow-lane FCB estimation is eliminated and more than 94% of the narrow-lane residuals are within 0.1 cycles for both the GPS and the BDS. The fixing percentage within 20 min is 46.3% for the GPS-only solution but is only 4.8% when using GPS + BDS with the traditional method. However, when the proposed strategy is used, the fixing percentage for GPS + BDS improves significantly to 91.7%.

In this paper, an improved method based on a mixture of Gaussian and quadrilateral functions is presented to process airborne bathymetric LiDAR waveforms. In the presented method, the LiDAR waveform is fitted to a combination of three functions: one Gaussian function for the water surface contribution, another Gaussian function for the water bottom contribution, and a new quadrilateral function to fit the water column contribution. The proposed method was tested on a simulated dataset and a real dataset, with the focus being mainly on the performance of retrieving bottom response and water depths. We also investigated the influence of the parameter settings on the accuracy of the bathymetry estimates. The results demonstrate that the improved quadrilateral fitting algorithm shows a superior performance in terms of low RMSE and a high detection rate in the water depth and magnitude retrieval. What’s more, compared with the use of a triangular function or the existing quadrilateral function to fit the water column contribution, the presented method retrieved the least noise and the least number of unidentified waveforms, showed the best performance in fitting the return waveforms, and had consistent fitting goodness for all different water depths.

Abstract: The spatial interpolation of property fields in 3D, such as the temperature, salinity, and organic content of ocean water, is an active area of research in the applied geosciences. Conventional interpolation methods have not adequately addressed anisotropy in these data. Thus, in our research we considered two interpolation methods based on a triangular prism volume element, as a triangular prism structure best represents directivity, to express the anisotropy inherent in geological property fields. A linear triangular prism interpolation is proposed for layered stratum that achieves a complete continuity based on the volume coordinates of the triangular prism. A triangular prism quadric interpolation (a unit function of a triangular prism spline with 15 nodes) is designed for a smooth transition between adjacent triangular prisms with approximately continuity, expressing the continuity of the entire model. We designed a specific model which accounts for the different spatial correlations in three dimensions. We evaluated the accuracy of our proposed linear and triangular prism quadric interpolation methods with traditional inverse distance weighting (IDW) and kriging interpolation approaches in comparative experiments. The results show that, in 3D geological modeling, the linear and quadric triangular prism interpolations more accurately represent the changes in the property values of the layered strata than the IDW and kriging interpolation methods. Furthermore, the triangular prism quadric interpolation algorithm with 15 nodes outperforms the other methods. This study of triangular prism interpolation algorithms has implications for the expression of data fields with 3D properties. Moreover, our novel approach will contribute to spatial attribute prediction and representation and is applicable to all 3D geographic information; for example, in studies of atmospheric circulation, ocean circulation, water temperature, salinity, and three-dimensional pollutant diffusion.

The Rational Function Model (RFM) is a widely used generic sensor model for georeferencing satellite images. Owing to inaccurate measurement of satellite orbit and attitude, the Rational Polynomial Coefficients (RPCs) provided by image vendors are commonly biased and cannot be directly used for high-precision remote-sensing applications. In this paper, we propose a new method for the bias compensation of RPCs using local polynomial models (including the local affine model and the local quadratic model), which provides the ability to correct non-rigid RPC deformations. Performance of the proposed approach was evaluated using a stereo triplet of ZY-3 satellite images and compared with conventional global-polynomial-based models (including the global affine model and the global quadratic model). The experimental results show that, when the same polynomial form was used, the correction residuals of the local model could be notably smaller than those of the global model, which indicates that the new method has great ability to remove complex errors existed in vendor-provided RPCs. In the experiments of this study, the accuracy of the local affine model was nearly 15% better than that of the global affine model. Performance of the local quadratic model was not as good as the local affine model when the number of Ground Control Points (GCPs) was less than 10, but it improved rapidly with an increase in the number of redundant observations. In the test scenario with 15 GCPs, the accuracy of the local quadratic model was about 9% and 27% better than those of the local affine model and the global quadratic model, respectively.

Object detection and reconstruction from remotely sensed data are active research topic in photogrammetric and remote sensing communities. Power engineering device monitoring by detecting key objects is important for power safety. In this paper, we introduce a novel method for the reconstruction of self-supporting pylons widely used in high voltage power-line systems from airborne LiDAR data. Our work constructs pylons from a library of 3D parametric models, which are represented using polyhedrons based on stochastic geometry. Firstly, laser points of pylons are extracted from the dataset using an automatic classification method. An energy function made up of two terms is then defined: the first term measures the adequacy of the objects with respect to the data, and the second term has the ability to favor or penalize certain configurations based on prior knowledge. Finally, estimation is undertaken by minimizing the energy using simulated annealing. We use a Markov Chain Monte Carlo sampler, leading to an optimal configuration of objects. Two main contributions of this paper are: (1) building a framework for automatic pylon reconstruction; and (2) efficient global optimization. The pylons can be precisely reconstructed through energy optimization. Experiments producing convincing results validated the proposed method using a dataset of complex structure.