Abstract
This paper proposes an augmented reality (AR) system for investigation of the pipelines for agricultural water use. While an industrial endoscope with a wide angle-view of video camera is one of the major tools for inspection of the pipelines instead of excavation, even if a severe damage is found and repair work is required, the above and underground situation around the damaged part of pipeline is not easily verified. In our scenario, in addition to the live camera images of the inner surface of the pipes, overlaying the images onto the inspector's view is effective to check the situation between above ground and inner pipe. However, the quality of mapped image of the wide-angle view image into development view or cylindrical surface is easily distorted due to the camera position and orientation. This paper proposes to introduce a depth camera to compasate the camera posuture changes. A depth camera can capture point cloud with normal vectors of the pipe. We estimate the direction from pipe relative to the camera coordinate, using the normal vectors of the point cloud. Correcting the camera posture based on the estimated relative pipe direction, mapping the endoscopic image onto a cylindrical 3D model of nominal pipeline size of the pipe visualizes the inside situation of the pipe without distortion. In our experiments by the proto-type system, as the endoscopy moves forward, the live texture-mapped 3D pipe model is properly overlaid onto the realtime video image of the exterior appearance of the scenery.
