An in-pipe internal defects inspection system based on the active stereo omnidirectional vision sensor

Abstract

To ensure the safety and serviceability of underground pipeline, it is essential to inspect and assess its structural and functional condition. Underground pipe inner-surface defect inspection is constrained by space accessibility. Thus, in this work we propose a comprehensive in-pipe inspection method based on active stereo omnidirectional vision. In our system, a crawling robot equipped with an active stereo omnidirectional vision sensor travels along the pipeline, taking panoramic images of the internal surface and laser streaks projected from an omnidirectional laser in real time. Afterwards, the internal surface panoramic images are disposed as follows: unwrapping, pre-processing and geometrical features extraction of the defect region, ultimately we classify the defects' category and degree. On the other hand, image sequences with laser streaks are used to extract 3D point cloud data of the inner-surface and calculate the deformation rate of the pipe, this part is not included in this paper. Experimental results show that our system is capable to achieve a qualitative and quantitative analysis of the in-pipe defects such as crack and corrosion, and provide a new approach for inspection of the in-pipe internal surface defects.