Crack Monitoring from Motion (CMfM): crack detection and measurement using cameras with non-fixed positions

Crack detection and measurement are standard procedures during inspections of infrastructures. Traditionally, these activities are only visually performed. To accomplish this, infrastructures must be closed and inspections are carried out at night to minimise the impact of the infrastructure downtime. The limited time and the length of the system make it impossible to inspect the infrastructure in detail, which increases the risk that cracks are not detected. In the last decades, image-based techniques such as Digital Image Correlation (DIC) have been used to measure deformation and cracks in image time series. Unfortunately, the main limitation is the requirement of collecting images with a fixed camera (fixed between each inspection/frame), which represents a strong limitation for long-term monitoring. Recently, inspections have also been carried out with Mobile Mapping Systems (MMSs) that can capture the scene using a set of geomatic sensors. Specifically, images collected with MMSs are used for finding cracks and monitoring their extent over time. Unfortunately, due to the limitations of standard DIC techniques, crack propagation cannot be measured with DIC using images collected from different points of view since the MMS camera position between the inspections differs.  In this work, we present a methodology (Crack Monitoring from Motion - CMfM) that integrates deep learning methods (Convolutional Neural Networks) with photogrammetric techniques for automatic detection and monitoring of cracks using a series of images collected with not fixed cameras [1]. Unlike conventional and image-based techniques, CMfM does not require fixed artificial targets and overcomes the DIC limitations of using fixed cameras, which opens up new possibilities for automatically monitoring crack propagation using images collected with MMS or standard cameras. The method can enable automatic monitoring of infrastructures, increasing the efficiency of the monitoring process and decreasing the risk that cracks are not found. The widespread adoption of CMfM can lead to significant improvements in the field of infrastructural monitoring and maintenance. Here, we present the results of crack detection and measurement during three-point bending tests on concrete beams. During the experiments, we used both fixed and not fixed cameras for collecting images and we processed the data with CMfM. We validated our methodology with comparisons with the standard DIC technique and local sensors such as Linear Variable Differential Transformers. We demonstrated that our algorithm can compute the crack width with an accuracy of a few hundredths of a millimetre compared to the adopted local sensor, demonstrating the possibility of measuring the crack evolution over time using non-fixed cameras. This work is part of the international TACK (Tunnel Automatic CracK Monitoring using Deep Learning) project [2].   [1] Belloni et al., Crack Monitoring from Motion (CMfM): Crack detection and measurement using cameras with non-fixed positions, Automation in Construction, Volume 156, 2023, 105072, ISSN 0926-5805, https://doi.org/10.1016/j.autcon.2023.105072. [2] Belloni et al., TACK project: tunnel and bridge automatic crack monitoring using deep learning and photogrammetry, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B4-2020, 741–745, https://doi.org/10.5194/isprs-archives-XLIII-B4-2020-741-2020, 2020.