Crack detection and measurement are standard procedures during infrastructure inspections all over the world. Traditionally, infrastructures are only visually inspected. To accomplish this, tunnels must be closed and inspections are carried out during the night to minimize infrastructure downtime. Also, the limited time and the length of the system make it impossible to accurately inspect tunnels which increases the risk that cracks are not detected. Recently, inspections have been carried out using semi-automatic methods where a mobile mapping equipment is used to capture the scene and reconstruct the 3D model (the socalled digital twin) of tunnels using geomatics sensors (visible and infrared cameras, LiDAR, IMU). Then, the digital twin and the images are manually analysed to find cracks and mark their extent. Due to the large amount of data, the method is time-consuming, inefficient and affected by errors. The aim of this work is to improve efficiency and accuracy of inspections.