Civil infrastructure defect assessment using pixel-wise segmentation based on deep learning

Abstract

Nowadays, the number of aging civil infrastructures is growing world-wide and when concrete is involved, cracking and delamination can occur. Therefore, ensuring the safety and serviceability of existing civil infrastructure and preventing an inadequate level of damage have become some of the major issues in civil engineering field. Routine inspections and maintenance are then required to avoid leaving these defects unexplored and untreated. However, due to the limitations of on-field inspection resources and budget management efficiency, automation technology is needed to develop more effective and pervasive inspection processes. This paper presents a pixel-wise classification method to automatically detect and quantify concrete defects from images through semantic segmentation network. The proposed model uses Deeplabv3+ network with weights initialized from pre-trained neural networks. The comparison study among the performance of different deep neural network models resulted in ResNet-50 as the most suitable network for applications of civil infrastructure defects segmentation. A total of 1250 images have been collected from the Internet, on-field bridge inspections and Google Street View in order to build an invariant network for different resolutions, image qualities and backgrounds. A randomized data augmentation allowed to double the database and assign 2000 images for training and 500 images for validation. The experimental results show global accuracies for training and validation of 93.42% and 91.04%, respectively. The promising results highlighted the suitability of the model to be integrated in digitalized management system to increase the productivity of management agencies involved in civil infrastructure inspections and digital transformation.

Image-based automatic multiple-defect detection of urban utility tunnel using UUTNet

“…Hyperparameters are model parameters that need to be confirmed before training deep learning models, and their selection will affect results of the defect detection (Chou and Nguyen, 2022). Currently, most machine learning or deep learning models select empirical parameters for hyperparameters setting and optimization (Liu and Chou, 2023; Savino and Tondolo, 2023), which is difficult to achieve the optimal global solution, and the defect detection accuracy still has room for improvement. Bayesian optimization algorithm (BOA) has been widely used in machine learning (Lu et al, 2020) and has been shown to outperform random search, grid search, and human search in hyperparameters optimization (Xia et al, 2017), therefore it is introduced into the PSPNet.…”

“…Guo (Guo, 2022) integrates binocular and monocular cameras with the encoder-decoder network for automatic detection, ranging, and quantification of cracks as well as characterization of crack patterns. Aiming at cracks and delamination of infrastructure, Savino (Savino and Tondolo, 2023) developed a universal semantic segmentation network adapted to different image qualities, resolutions, and backgrounds. They introduced the ResNet50 module in the Deeplabv3+ network, utilized experience to set hyperparameters, and finally achieved a validation accuracy of 91.04%.…”

Critical Infrastructure Asset Imaging Pipeline

“…Models have been developed for automated detection, identification, and segmentation of infrastructure elements, such as segmenting the individual elements that can comprise a power utility pole [13], [9], [14], [15]. Deep learning is also being used for the detection and segmentation of defects in critical infrastructure including segmentation of cracks in concrete structures [16], [17], [18], or measurement of the inclination of utility poles [15].…”

Deep Learning-Based Steel Bridge Corrosion Segmentation and Condition Rating Using Mask RCNN and YOLOv8

“…The manual inspection of the bridge is always time-consuming and the result is inconsistent, subjective, and largely dependent on the experience and judgment of the inspector [2,3]. In addition, traditional bridge inspection is costly and associated with safety risks for the inspection crew and public.…”