Pavement crack detection from images is a challenging problem due to intensity inhomogeneity, topology complexity, low contrast, and noisy texture background. Traditional learning-based approaches have difficulties in obtaining representative training samples. We propose new unsupervised multi-scale fusion (MFCD) algorithm that does not require data. First, we develop windowed minimal path-based method extract the candidate cracks image at each scale. Second, find correspondences across...

Detection of subsurface defects is important for maintaining runway structural health and reliability. A potential solution to employ a robot equipped with Ground Penetrating Radar (GPR) perform scanning. To automate the inspection process, we develop defect detection algorithm which deep learning that fuses 2D planar features in each panel GPR B-scans 3D voxel-wise C-scan robustly detect regions even presence significant noises. Named as GPR-RCNN, have tested our real airport data collected...

In order to improve the accuracy and robustness of existing automated crack detection methods, a fully convolutional neural network for pixel-level based on densely connected deeply supervised is proposed. First, layers are applied enhancing propagation reuse features. Then, modules designed make extract more significant features through multi-scale levels. Finally, feature maps from different scales fused achieve complementarity at addition, class-balanced cross-entropy loss function...

We propose a novel subsurface pipeline mapping and 3D reconstruction method by fusing ground-penetrating radar (GPR) scans camera images. To facilitate the simultaneous detection of multiple pipelines, we model GPR sensing process prove hyperbola response for general scanning with nonperpendicular angles. Furthermore, fuse visual localization outputs, encoder readings to classify hyperbolas into different groups. extensively apply J-linkage maximum likelihood estimation error analysis...

<title>Abstract</title> Reranker and retriever are two important components in information retrieval. The typically adopts a dual-encoder model, where queries documents separately input into pre-trained models, the vectors generated by models used for similarity calculation. reranker often uses cross-encoder concatenated query-document pairs model to obtain word similarities. However, lacks interaction between due its independent encoding, while requires substantial computational cost...

Abstract Deep reinforcement learning (DRL) can be used to extract deep features that incorporated into systems enable improved decision-making; DRL therefore also for managing stock portfolios. Traditional methods cannot fully exploit the advantages of because they are generally based on real-time quotes, which do not have sufficient making comprehensive decisions. In this study, in addition we introduced financial indices as additional features. Moreover, Markowitz mean-variance theory...

Detection of both surface and subsurface defects is a vital task for maintaining the structural health reliability airport runways. We report automated data collection analysis runways based on our novel robotic system, which employs camera GPR (Ground Penetrating Radar) to inspect conditions, respectively. To perform analysis, we propose crack detection algorithm images, defect method with data. Additionally, create composite global view large runway span, camera/GPR sequence from robot...

The detection and restoration of subsurface defects are essential for ensuring the structural reliability airport runways. Subsurface inspections can be performed with aid a robot equipped Ground Penetrating Radar (GPR). However, interpreting GPR data is extremely difficult, as usually contains severe clutter interference. In addition, many different types present similar features in B-scan images, making them difficult to distinguish. Consequently, this makes later maintenance work harder...

To facilitate scene understanding and robot navigation in a modern urban area, we design multilayer feature graph (MFG) based on two views from an on-board camera. The nodes of MFG are features such as scale invariant transformation (SIFT) points, line segments, lines, planes while edges the represent different geometric relationships adjacency, parallelism, collinearity, coplanarity. also connects corresponding 3D coordinate system. Building SIFT points is constructed using fusion which...

Efficient scheduling is of great significance to rationally make use scarce satellite resources. Task clustering has been demonstrated realize an effective strategy improve the efficiency scheduling. However, previous task static. That is, it integrated into in a two-phase manner rather than dynamic fashion, without expressing its full potential improving performance. In this study, we present adaptive Simulated Annealing based algorithm aggregated with (or ASA-DTC for short) observation...

Nowadays the phenomenon of flight delays and cancellations is becoming more serious. Flight not only waste transportation resources, but also affect passengers' travel plans, which cause an increase in passenger discontent complaint rates[2]. The dissatisfaction distrust airlines seriously damage airlines' corporate reputation then loyalty. Therefore, based on information 5 million data United States 2016, we predicted four classification algorithms: logistics regression, support vector...

In this article, we report system and algorithmic developments for a sensing suite comprising camera ground penetrating radar (GPR) with wheel encoder designed both surface subsurface infrastructure inspection, which is multimodal mapping task. To fuse different sensor modalities properly, solve novel GPR-camera calibration problem synchronization-challenged fusion problem. First, design rig, model the GPR imaging process, introduce mirror to obtain joint coverage between GPR, employ...

We report system and algorithmic development for a sensing suite comprising multiple sensors both surface subsurface transportation infrastructure inspection focusing on multi-modal mapping inspection. The contains camera, ground penetrating radar (GPR), wheel encoder. design the propose data collection scheme using customized artificial landmarks (ALs). use ALs to synchronize two types streams: camera images that are temporally evenly-spaced GPR/encoder spatially evenly-spaced. also employ...

The fluctuation of stock price is a major concern for investors. In order to provide investors with decision-making suggestions buying or selling stocks, we use big data computing platform-Spark perform binary classification prediction stocks' rise and fall according the related information in trading day, compare four algorithms: naïve bayes, random forest, decision tree, logistic regression from metrics: run time, accuracy, AUC(Area under receiver operating characteristic curve) PR(Area...

Buildings consume around 40% of overall energy in the world. Planar mirror detection problem (PMDP) arises when surveying reflective building surface for retrofit. PMDP is also important collision avoidance robots navigate close to highly glassy walls. Our approach uses two views from an on-board camera. First, we derive geometric constraints corresponding real-virtual features across views. The include 1) normal as a function vanishing points lines connecting feature point pairs and 2)...

We propose a novel subsurface pipeline mapping method by fusing Ground Penetrating Radar (GPR) scans and camera images. To facilitate the simultaneous detection of multiple pipelines, we model GPR sensing process prove hyperbola response for general scanning with non-perpendicular angles. Furthermore, fuse visual localization outputs, encoder readings to classify hyperbolas into different groups. extensively apply J-Linkage maximum likelihood estimation improve algorithm robustness accuracy....

The low-level line segment features have low accuracy as they are more easily affected by the noise and differeent detectors. Furthermore, is not a good feature for matching across multiple views when we need to finish 3D reconstruction. However, robust noise. In this paper, kind of feature, ideal line, defined optimally estimated clustering segments under RANSAC framework. physical experiments carried out verify proposed estimation method.

Ground penetrating radar (GPR) is widely used for detecting airport runway subsurface defects. The trailing interference in the GPR data disguises defect responses, which seriously affects accuracy of detection. To tackle challenge defects detection under real scenarios, a new multi-modal fusion network referred to as M <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> FNet proposed. Based on premise that signal highly similar across all...

Traditional manual visual inspection of the airplane surface has disadvantages low efficiency, poor safety, and incapability full coverage. To address these problems, unmanned ground vehicle (UGV) is employed to perform coverage automatically. Viewpoint generation for UGV, which key high rate quality, main concern this article. improve photogrammetric requirements including equidistance, front view, overlap are introduced. Furthermore, many practical constraints, such as safe distance,...

In this paper, several similarity measures of fuzzy soft sets are introduced based on four types quasi-metrics. Meantime, an illustrative example is also given.

Highly accurate sensor extrinsic calibration is critical for data fusion from multiple sensors, such as camera and Inertial Measurement Unit (IMU) suit. A pre-calibrated extrinsics, however, may no longer be due to external disturbances, e.g., vehicle vibration, which will lead significant performance deterioration of autonomous vehicles. Existing approaches rely on online recalibration at a fixed frequency regardless whether the extrinsics have actually been changed or needed,...