Infrastructure is deteriorating rapidly, necessitating improved efficiency in inspection work. Particularly in expressway tunnels, the introduction of mobile infrastructure inspection vehicles is anticipated, as they can be monitored while moving. Regular monitoring is crucial because the tunnel lining surface is attached to components such as jet fans, lighting, and cables, which are held in place by screws and nuts. Since these attachments are located overhead from the perspective of a moving vehicle, there is a significant risk of them falling. Close-up inspections are essential to prevent this, but conducting them frequently is challenging due to the need for road closures.
In this study, we explored the feasibility of stereo matching using images captured by a single camera mounted on a moving infrastructure inspection vehicle. Stereo matching involves aligning feature points of images taken from two different viewpoints and estimating the distance to the object based on the disparity between the images. Normally, stereo cameras are not designed for high-speed motion, such as at 100 km/h. In this paper, a series of 2D images taken from a developed motion blur compensation system while moving are used. This system aims to enable stereo matching from a single camera at high speeds, developing essential technology for infrastructure monitoring. In the experiment, flat and uneven objects were used as image subjects, attached to a linear actuator for movement. The motion-blur compensated images were reconstructed as 3D data, and depth accuracy was evaluated. Finally, this accuracy was applied to analyze data captured in a tunnel at 100 km/h, estimating the depth of wires, metal plates, and cables.
As a future development, stereo matching with multiple traveling images will be used to deal with the lateral direction also to be applied for the motion in any direction, and this method will be adopted as the part of the infrastructure management flow.