Bridge decks deteriorate over time as a result of deicing salts, freezing-and-thawing, and heavy use, resulting in internal defects. According to a 2006 study by the American Society of Civil Engineers, 29% of bridges in the United States are considered structurally deficient or functionally obsolete. Ground penetrating radar (GPR) is a promising non-destructive evaluation technique for assessing subsurface conditions of bridge decks. However, the analysis of GPR scans is typically done manually, where the accuracy of the detection process depends on the technician's trained eye. In this work, a framework is developed to automate the detection, locailzation, and characterization of subsurface defects inside bridge decks. This framework is composed of a fractal-based feature extraction algorithm to detect defective regions, a deconvolution algorithm using banded-ICA to reduce overlapping between reflections and to estimate the depth of defects, and a classification algorithm using principal component analysis to identify main features in defective regions. This framework is implemented and simulated using MATLAB and GPR real scans of simulated concrete bridge decks. This framework, as demonstrated by the experimental results, has the following contributions to the current body of knowledge in ground penetrating radar detection and analysis techniques, and in concrete bridge deck condition assessment: 1) developed a framework that integrated detection, localization, and classificationof subsurface defects inside concrete bridge decks, 2) presented a comparison between the most common fractal methods to determine the most suitable one for bridge deck condition assessment, 3) introduced a fractal-based feature extraction algorithm that is capable of detecting and horizontally labeling defective regions using only the underlying GPR B-scan without the need for a training dataset, 4) developed a deconvolution algorithm using EFICA to detect embedded defects in bridge decks, 5) introduced an automated identification methodology of defective regions which can be integrated into a CAD system that allows for better visual assessment by the maintenance engineer and has the potential to eliminate human interpretation errors and reduce condition assessment time and cost, and 6) presented an investigation and a successful attempt to classify some of the common defects in bridge decks.

Structural inspection and maintenance of bridges are essential to improve the safety and sustainability of the infrastructure systems. Visual inspection using non-equipped eyes is the principal method of detecting surface defects of bridges, which is time-consuming, unsafe, and encounters inspectors falling risks. Therefore, there is a need for automated bridge inspection. Recently, Light Detection and Ranging (LiDAR) scanners are used for detecting surface defects. LiDAR scanners can collect high-quality 3D point cloud datasets. In order to automate the process of structural inspection, it is important to collect proper datasets and use an efficient approach to analyze them and find the defects. Deep Neural Networks (DNNs) have been recently used for detecting 3D objects within 3D point clouds. PointNet and PointNet++ are deep neural networks for classification, part segmentation, and semantic segmentation of point clouds that are modified and adapted in this work to detect surface concrete defects. The research contributions are: (1) Designing a LiDAR-equipped UAV platform for structural inspection using an affordable 2D scanner for data collection, and (2) Proposing a method for detecting concrete surface defects using deep neural networks based on LiDAR generated point clouds. Training and testing datasets are collected from four concrete bridges in Montréal and annotated manually. The point cloud dataset prepared in five areas, which contain more than 51 million points and 2,572 annotated defects in four classes of crack, light spalling, medium spalling, and severe spalling. The accuracies of 75% (adapted PointNet) and 79% (adapted PointNet++) in detecting defects are achieved in binary semantic segmentation.