UAV data acquisition method for transportation tunnel inspection

Loading...
Thumbnail Image
Files
AV.pdf(516.15 KB)
Accepted Version
Date
2023
Authors
Zhang, Ran
Li, Zili
Journal Title
Journal ISSN
Volume Title
Publisher
Published Version
Research Projects
Organizational Units
Journal Issue
Abstract
In a large-scale underground transportation network, miles of tunnel linings require regular inspection and assessment. In civil / geotechnical industry, common routine tunnel maintenance still highly relies on labour-intensive visual inspection, manual data acquisition and subjective assessment, which leads to significant cost for a large-scale tunnel network over many miles. In this paper, the tunnel surface texture is acquired using a commercial DJI Mavic2 UAV customized by DJI MSDK automation. Due to limited GPS signal in an underground tunnel, optical flow method is used for localization. In railroad tunnels without a priori maps, the adaptive flight procedure allows the UAV to determine the tunnel axial direction and adjust the UAV orientation and position relative to the tunnel in real time. Compared to manual acquisition and manual control, the UAV automatic data acquisition procedure allow a larger inspection scope, it has higher stability and greater efficiency at a lower cost, and can create a 3D visual model of the entire tunnel in the simple geometry of railway tunnels. Compared to more sophisticated programs, the light-weight platform can be easily compiled on any DJI Mavic 2 model with low computation power and adapt to similar underground environments without the need of pretrained datasets. In future studies, the obtained tunnel images can be used to train convolution neural networks (CNN) for automatic cracks & leakage detection and tunnel condition assessment.
Description
Keywords
Structural health monitoring , Image acquisition , UAV , Point clouds
Citation
Zhang, R. and Li, Z. (2023) 'UAV data acquisition method for transportation tunnel inspection', Proceedings of the Fourth International Symposium on Machine Learning and Big Data in Geoscience, Cork, Ireland, 29 August - 1 September. Extended Abstract 88, pp. 212-214.
Link to publisher’s version
Copyright
© 2023, the Authors.