Answers are grounded in this paper and cite their sources.
Surface cracks on buildings, natural walls and underground mine tunnels can serious structural integrity issues that threaten the safety of the and people in the environment. Timely detection and monitoring of are crucial to managing these risks, especially if the systems can be highly automated through robots. Vision-based crack detection algorithms deep neural networks have exhibited promise for structured surfaces such walls or civil engineering tunnels, but little work has addressed highly environments such as rock cliffs and bare mining tunnels. To this challenge, this paper presents PointCrack3D, a new3D-point-cloud-based crack detection algorithm for unstructured surfaces. The comprises three key components: an adaptive down-sampling method that sufficient crack point density, a DNN that classifies each point as or non-crack, and a post-processing clustering method that groups crack into crack instances. The method was validated experimentally on a new natural rock dataset, comprising coloured LIDAR point clouds spanning than 900 m^2 and 412 individual cracks. Results demonstrate a crack rate of 97% overall and 100% for cracks with a maximum width of more 3 cm, significantly outperforming the state of the art. Furthermore, for-validation, PointCrack3D was applied to an entirely new dataset acquired different locations and not used at all in training and shown to detect 100% its crack instances. We also characterise the relationship between detection, crack width and number of points per crack, providing a foundation which to make decisions about both practical deployments and future directions.
No takes yet. Share an insight, caveat, or question.
Azhari et al. (2021) studied this question.