Automatic visual detection of lateral junctions in sewers

Attention is focused on the automatic detection of lateral connections in closed-circuit television images obtained during the survey pass of an Insituform renovation process with regard to non-man-entry sewers. It is suggested that computer vision, in the form of edge detection and application-specific selection criteria, provides a practical and economically viable method of lateral detection. The illumination conditions under which these techniques are applied are, however, crucial; it is demonstrated that detection is greatly eased by the use of reflected, rather than direct illumination of the region under consideration. The incorporation of image enhancement prior to image analysis provides further system robustness. The proposed method has been implemented and shown to detect lateral junctions in images obtained both from laboratory and field environments.