The adoption of unmanned aerial vehicle (UAV) has grown over the last few years in different industries. Common bridge inspection procedures are extensively time consuming, highly subjective to visual interpretation, and associated with safety risks. The primary goal of this paper is to develop a sufficiently accurate and cost-effective crack detection framework using a UAV mounted red, green, and blue (RGB) camera for data collection and deep convolutional neural networks (CNN) on a single board computer for bridge inspection. A concrete crack data set is created using a RGB sensor mounted on a UAV, named UAVCON. Single shot multibox detector (SSD) and region-based convolutional neural networks (R-CNN) meta-models are deployed in TensorFlow framework for benchmarking on the data set. The proposed detection model was able to detect concrete cracks on bridge elements with 93.4% accuracy on UAVCON.
Deep Convolutional Neural Networks for Crack Detection Using a Cost-Efficient Single Board Computer-Based Unmanned Aerial Vehicle (UAV) Platform for Bridge Inspection
International Conference on Transportation and Development 2022 ; 2022 ; Seattle, Washington
2022-08-31
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
BASE | 2021
|