Vehicle-mounted multichannel ground penetrating radar (MC-GPR) facilitates acquisition of volume images of subsurface structures, but inconsistencies arise due to different transmitted wavelets from each antenna. This paper introduces a methodology, Reflectivity-Consistent Sparse Blind Deconvolution (RC-SBD), that estimates transmitted wavelets and stationary clutter for each antenna, thereby calibrating GPR volume images. Previous Sparse Blind Deconvolution was achieved by alternating minimization, producing a single transmitted wavelet and sparse ground reflectivity. RC-SBD utilizes an assumption of subsurface reflectivity smoothness in the horizontal direction, expressed via a total variation regularization term. This allows a wavelet to be derived for each channel. Stationary clutter variables, such as reflection from the vehicle itself and direct waves, are integrated into the model for simultaneous estimation. The objective function incorporates several ℓ2 and ℓ1 regularization terms and is solved using the Split Bregman algorithm augmented by a gradient method. Hyperparameters are determined through Bayesian optimization, aiming to maximize kurtosis of the frequency domain calibrated volume image. The proposed methodology was validated with synthetic data, demonstrating accurate wavelets estimation and significant denoising of volume image. Real-world data application revealed substantial enhancements in the channel depth cross section, visualizing responses of structures such as rebar and steel plates. Furthermore, calibrated image remained stable across diverse datasets, including earthwork and bridge sections, indicating the estimates’ versatility.

Bridge finger type expansion joints are widely used because of their durability. However, they are susceptible to damage from fatigue and corrosion, and making the development of efficient damage detection methods crucial for preventing serious accidents. This study proposes a framework for automatic damage detection by simply driving in a vehicle, consisting of two steps: (1) detection of expansion joints by applying object detection to smartphone video, and (2) calculation of damage scores by passing sound recorded by an on-vehicle microphone. YOLOv5 object detection successfully detected 99.8% of expansion joints in the first step. Regarding the second step, experiments were conducted using one intact and four specimens modeling incremental damage. Finally, the damage score was proposed to quantify the sound of contact between among damaged structural members. After applying the proposed method to expansion joints that are in service, those identified as needing urgent replacement (Urgent Replacement Needed or U.R.N specimens) scored in the top 0.1% of the distribution of scores, indicating low false positives.

Smartphone-mounted handheld GPR are mainly used for detecting rebar and are attracting attention in the field of infrastructure maintenance. However, the information obtained by electromagnetic wave radar is the round-trip propagation time of electromagnetic waves from the transmitting antenna to the receiving antenna, which requires estimation of the relative permittivity in the medium to be converted into distance information. In this study, we propose a method for estimating relative permittivity and depth that takes advantage of curve reactions generated from any reflective object, including rebar and cracks. First, we confirmed that the relative permittivity can be estimated with practicable accuracy by curve fitting, in which theoretical and measured curves are observed. Next, we proposed an algorithm that automatically extracts the range to be fitted, making it possible to obtain the relative permittivity and the depth to be estimated simultaneously. As a result, we succeeded in estimating the relative permittivity with high accuracy on the millimeter-order of less than 5% error in the 0.1 second range.

Smartphone-Mounted Handheld GPR are used to detect rebar, but quantitative detection of damage is difficult due to small reflectance. In this study, we propose an algorithm to quantitatively estimate damage thickness from received waveforms obtained from electromagnetic radar. A simple estimation method that separates peaks from above and below the crack from the time waveform is extremely difficult due to the presence of subtractive interference and side lobes. Therefore, we focus on the amplitude spectrum with the phase information dropped and estimate the damage thickness by pattern matching with a rigorous theoretical spectrum that takes multiple reflections into account. Rough damage thickness is determined by first combining multiple centroids of amplitude intensity and low-frequency components, and then spectral pattern matching is performed within the gating and bandwidth ranges set for each region, which also contributes to improving processing speed. As a result, damage thicknesses from 2 mm to 180 mm could be quantitatively estimated with a single algorithm in the 0.001 second range, confirming the practical feasibility of this algorithm.

A completely automatic and accurate detection algorithm for the delamination on tunnel concrete lining surfaces from laser 3D point cloud data is proposed. A Mobile Mapping System (MMS), which mounts laser sensors and a positioning system, is utilized to measure the geometry of tunnel lining surfaces highspeed. The proposed algorithm consists of 4 steps: removal of tunnel profile components, detection of peaks of anomalies, localization of anomaly areas, classification of delamination and appendages. On tunnel linings, there are many appendages such as cables, lights, signs, and water guides to mask the features of delamination. In the article, a novel 3D feature was introduced to realize the accurate classification. An automatic SVM algorithm was developed using real tunnel lining data and manual inspection results, showing an accurate delamination map.

Aging of infrastructures has been a worldwide issue, and cost saving by shifting to preventive maintenance is urgent. Especially, damage detection of concrete bridge decks is one of the most important subjects, because of the significant repair costs due to its complicated structure. Rebars are the fundamental components of bridge decks, and they often become the trigger of bridge decks’ damages. Previous researches have been focused on detecting the locations of rebars in cross section images acquired by single-channel ground penetrating radar (GPR), however, no research has reproduced 3D rebar mesh arrangement from radar volume images acquired by multi-channel GPR.This paper proposes a method that reproduces 3D rebar mesh that contains the data of vertex location and reflection time from radar volume images. Real scale bridge deck specimens were created in this study and reflections of electromagnetic waves were observed utilizing an on-vehicle GPR. In the proposed method, 3D filtering based on the 3D DFT theory for the noise reduction was applied. Also, automation in detection of the rebar depth was achieved focusing on edges of the images. As a result, 3D rebar meshes were successfully reproduced with appropriate distribution rebar spacing. Also, the method was applied to radar data acquired from a bridge in service. The proposed method effectively functioned for in-service bridge data, and automatically reproduced the mesh of 21m length.

On-vehicle multi-channel Ground Penetrating Radar is attracting much attention in infrastructure maintenance field. Besides the key benefits of highspeed 3D subsurface data acquisition, basic problems such as data discontinuity due to use of multiple antennas and sparse observation that arises aliasing have been a problem. This study proposes easily applicable but effective method to solve these problems. For the first issue, calibration in each channel was done by improved Sparse Blind Deconvolution (SBD) method proposed in previous research. It was found that disuse of low-intensity frequency band provides better estimation of transmitted waves. For the second issue, super resolution method was proposed. It interpolates zeros in between sparsely observed data and then apply synthetic aperture processing. Zero insertion is an operation that restores high frequencies, which makes it possible to restore frequency components above the Nyquist frequency under certain conditions. The method was confirmed to be effective on real data.

The increasing length of the subsurface pipe causes overlapping, accumulating, and sometimes the old pipe layout is not available. Consequently, accidents, damages, time delay and financial losses are occurred during construction of new structures or expanding the roadway or installing of new pipes. Therefore, the depth, diameter, material of the existing pipe as well as the map of the pipe are indispensable to know for appropriate construction planning and development work. Using this algorithm and genetic algorithm optimization, the radius of the field pipe was assessed with 83, 67 and 89% accuracy, the depth was estimated with 95, 95 and 98% accuracy considering the effect of the pipe radius. The effect of the pipe radius should be considered to calculate the pipe depth with higher accuracy. The material of the field pipe was successfully determined using the evaluated relative permittivity of the pipe. A 3D map of the field pipe was developed by applying the tracing algorithm and linear regression to the estimated depth.

Road cracks are an important concern of administrators. Visual inspection is labor-intensive and subjective, while previous algorithms detecting cracks from optical camera images were not accurate. Furthermore, the actual length and thicknesses of a crack cannot be estimated only from images. Light Detection and Ranging (Lidar) is a standard feature introduced on the latest smartphones. In this research, for completely automatic, accurate and quantitative road crack evaluation using smartphones, an up-to-date segmentation technique, U-Net with morphology transform adopting data augmentation was proposed. Lidar 3D point cloud data of smartphones is linked to color data obtained from cameras. By registering images to Lidar data, geometrical relationships were estimated to calculate the length and thicknesses. The proposed algorithm was validated by a standard database of road cracks and dataset constructed by the authors, showing 95% length accuracy and 0.98 coefficient of determination for thickness estimation irrespective of various crack shapes and asphalt pavement color patterns.