Damage detection in subway tunnels is important for maintenance and is very labor intensive and time consuming. In recent years, machine vision has been applied to surface damage detection because of its noncontact tracking and recognition of surface information. Based on machine vision technology, a large number of tunnel detection systems have been developed, but both high detection efficiency and accuracy cannot be achieved at the same time with current subway tunnel systems. Additionally, the development of a system postprocessing platform has been lagging; thus, it has been difficult to meet the time limit and tremendous detection workload of China's subway tunnels. Therefore, more powerful detection equipment is needed. To obtain high-quality tunnel lining surface images during high-speed detection, in this study, subway tunnel rapid detection equipment is designed based on area-scan charge-coupled device (CCD) cameras. In addition, considering the quality of image acquisition, the tunnel vision system and light compensation system are optimized. For reliable mileage information, a multilocation system for locating damage is proposed. Furthermore, a three-level physical vibration reduction method is designed for reducing the vibration influence of maintenance trains that run during detection. The software system is developed with functions for image fusion, image preprocessing, and damage identification and a data platform. A deep learning algorithm is used to identify the damage features of the collected images. The powerful data platform provided by the software system can help tunnel managers view tunnel damage information and detection results in real time. Finally, field detection is undertaken to verify the efficiency and accuracy of the equipment, which shows that the developed detection equipment is suitable for surface damage detection in subway tunnels.
