With the mileage and density of high-speed railway growing rapidly, the tunnels are prone to various defects, which make the tunnels require regular inspection. Due to the short operation window of tunnels, traditional inspection methods are inefficient, unable to complete rapid inspection in a short time and thus affecting the normal operation of trains and timely rectification of defects. Based on the background, targeting the lining leakage, we have developed a rapid inspection device. And the device is used for on-site testing of more than 1000 km of high-speed tunnels in the southwest of China. The inspection speed of the device can reach 80 km/h and the pixel accuracy of the images is all higher than 0.4 mm/pixel, achieving efficient and rapid inspection. Additionally, based on the YOLOv5 model and after analyzing the feature of railway tunnel lining leakage, a rapid recognition model for lining leakage has been established. The YOLOv5 model is trained using the SimSiam model and images collected from on-site measurements with the device. When the model operates at a real-time detection speed of 32 FPS, the optimal detection accuracy can reach 0.722, showing good results, thus effectively promoting the intelligent operation and maintenance of high-speed railway tunnels.