Automatic analysis of visual data in submarine pipeline inspection

The aim of this work is to present an automatic system for analysis and interpretation of visual data in submarine pipeline inspection and to discuss its use in a navigation and inspection system for underwater applications. The system is conceived for working on-line on the images collected by cameras mounted on an unmanned remotely operated vehicle (ROV) that moves along the pipeline. The main task of the system consists in supporting the operator's activity in navigating the ROV and in detecting structural elements of the pipeline (anodes, gravel heaps, reference bands) and possible alien objects in its vicinity. The analysis of the data is performed by implementing image processing operations (filtering, contour extraction, feature enhancement) on a dedicated board. Then, the system employs a feature based technique for recognising the pipe's profile and for detecting structural elements and objects. The capabilities of the system have been evaluated, using archival images, in situations that emulate real working conditions. Thanks to the exploitation of the hardware architecture and the use of an efficient dynamic process in handling the information extracted from the image flow, the system reaches satisfactory performances in real time operation. The information gathered by analysing and interpreting the visual data can be used for automatic guidance and inspection purposes by integrating the system into a larger architecture. In this way, one obtains a semi-automatic navigation and inspection system for ROVs whose characteristics are described and discussed.