Detecting Leakage Water of Shield Tunnel Segments Based on Mask R-CNN

As a common shield tunnel problem, segment water leakage has a great impact on durability and applicability of tunnels. Therefore, the detection of leakage water is an important task for maintenance of shield tunnel. However, the traditional detection method is based on conventional image processing technology which has low accuracy, poor robustness, being difficult to meet the needs of actual engineering. In this paper, we proposed a leakage detection model based on deep learning Mask R-CNN through transfer learning with small-scale dataset. Our study shows that the proposed detection model is superior to the traditional image recognition algorithm in terms of accuracy, precision and recall.