3D Reconstruction of Sewer Shafts from Video

In this paper we propose a robust approach for automatic 3d model acquisition of sewer shafts from survey videos. Images are captured by a specific camera setup which is composed of a downward-looking fisheye-lens camera while lowering it into the shaft. Additionally, an inertial sensor measures rotation around the viewing axis. Our approach is based on Structure from Motion adjusted to the constrained motion and scene geometry, and measures the profile of the shaft using robust 2d shape recognition techniques. Global bundle adjustment is avoided by applying a simple and fast geometric correction of the computed 3d reconstruction which is also capable of handling inaccuracies of the intrinsic camera calibration parameters. An implementation of our method has been evaluated extensively with real data. Furthermore, we have proposed modifications of our so far off-line implementation to approach real-time reconstruction which can be applied during on-site inspection.