结合LiDAR强度数据与深度学习的盾构隧道渗漏水定量检测方法研究

针对传统盾构隧道渗漏水检测技术自动化程度低且缺乏渗漏水定量分析等问题，本项目以移动LiDAR系统获取的盾构隧道点云强度数据为研究对象，以深度学习方法为指导，通过分析移动LiDAR点云强度影响机制，研究盾构隧道特殊环境下移动LiDAR点云强度数据统一改正方法；在盾构隧道点云强度改正基础上，通过研究稀疏标注样本下盾构隧道强度图像渗漏水自动识别；分析盾构隧道表面含水量与校正后强度数据之间的关系，构建协同训练下融合多种深度卷积神经网络结构的盾构隧道渗漏水表面强度-含水量反演模型，研究盾构隧道强度图像渗漏水表面含水量定量分析方法；研究形成LiDAR强度数据与深度学习方法支持下盾构隧道渗漏水定量检测理论与方法，以实现盾构隧道渗漏水自动识别以及盾构隧道渗漏水表面含水量反演。预期研究成果对LiDAR技术在隧道病害检测领域有着重要的应用价值，同时在影像目标识别、目标含水量估算等多个领域也有着广泛应用前景。

Water leakages are very important signals that characterize the serious potential structural damages or flaws in shield tunnels. Automatic, timely, and quantitative detection of water leakages is of great significance to the safe operation and maintenance for shield tunnels. However, the traditional methods for water leakages detection are highly limited by the confined spaces and dim light conditions in shield tunnels, and also have been proved to be low-automation, and lack of quantitative analysis. As a new data acquisition technology, LiDAR can obtain high-density and high-precision 3D coordinates and record a co-located intensity value of shield tunnel , has shown its potentials in shield tunnel detection field. .This research aims to explore methodology for high precision water leakages recognition and water content inversion from mobile LiDAR intensity data based on deep learning. To do this, the unified correction method of mobile LiDAR point cloud intensity data under special environment of shield tunnel is explored based on the effect mechanism of mobile LiDAR point cloud intensity, then the intensity images of the shield tunnels were generated after intensity correction. secondly, under sparse labeled samples, the method of water leakages automatically recognition is studied through intensity image semi-supervised semantic segmentation. Finally, to analysis the relationship between the shield tunnel surface water content and the corrected intensity data, the water content inversion model of shield tunnel water leakage, combined with multiple deep convolutional neural network structures, is constructed by collaborative training. And the quantitative analysis method for water leakage region in the shield tunnel intensity image is studied. With the proposed and studied methods, the achievements are expected to have a wide range of applications, from the field of tunnel disease detection, image recognition to water content estimation of different target.