In this contribution we will present how visual odometry has been applied to a novel family of inspection robots for large sewer pipes in order to achieve a highly accurate, automated inspection. The inspection robots have been developed especially for use in the new Emscher Sewer, which will replace the Emscher River as the means for transporting wastewater through the densely-populated Ruhr-Rhein Area in Germany. The sewer will be a single-pipe system along much of its length. Normally a sewer of this length (over 57 km) is designed as a two-pipe system, so that while one pipe is in operation, the parallel pipe can be inspected, cleaned and repaired. While the cost-saving potential of a one-pipe system is very large, it also requires that damage such as cracks and corrosion be located at an early stage and compared over time in order for long-term correctional steps, e.g. construction of a replacement pipeline, to be taken. A much higher quality of inspection compared to current video sewer inspection systems was required in order to track small changes over time. Cracks with a width of less than 1 mm need to be located and compared with previous inspections over a course of many years. Precise and accurate information about the pose of the robots during the inspection process is needed in order to successfully create a 3-D representation of the inner pipe surface. Given that the distance between entrance shafts range up to 600 meters, standard techniques for determining the position of the robots along the length of the pipe by such as wheel-based odometry or measuring the length of the cable rolled from the winch are not accurate enough to achieve the desired inspection accuracy. The use of visual odometry as presented generates information about the position of the robot that is accurate enough for the goals of the inspection. This method constitutes the backbone of the inspection process, in which data from multiple sensors and cameras are fused to create a complete image of the pipe and allow for a highly accurate inspection and automated diagnosis of the state of the pipeline.
展开▼