Image processing and analysis as a diagnostic tool for an underground infrastructure technical condition monitoring

Difficult operating conditions of underground infrastructure create the need for regular inspection and monitoring of damage growth. In many cases, direct access to the surface of the lining can be very difficult. A separate problem is the large surface area and time of inspection, often exceeding the capabilities of human perception. In this situation, the easy way can be in the form of a video stream surface image recording and its automatic analysis in an on-line or off-line mode. Diverse structure shapes, spatial conditions and lighting conditions require the use of separate algorithms adapted to the specificities of detected problems. The paper presents an overview of possible solutions, algorithms and criteria for detecting different damage kinds and proposals for numerical description of their size. There have been shown examples for concrete and for brick lining surface analysis. Dripstones as the first symptoms of the initial phase of the damage can be identified on the basis of the color difference (related to the mineral contents). Cracks can be identified as elongated dark lines of the mostly irregular character (distinct from man-made components with a relatively simple and regular shape). Larger cavities can be identified on the basis of various textural parameters (detached fragments have typically much larger granularity than the remaining concrete lining; in the case of brick lining there is a breach of dimensional regularity of bricks and joints pattern). The most convenient way of detected damage presentation is a color enhancement and their overlay onto the true monochrome image plane and automatic measurement of such damage parameters as length, width or area. This—facilitating human interpretation— enhancement of the true image is called Augmented Reality (unlike Virtual Reality— generated entirely by the computer system). In many cases, processing and analysis of the image can be significantly hampered by the poor quality of the input video stream (poor and uneven illumination, glare on the wet surface, falling droplets, haze). Therefore, it is important to use appropriate methods of initial image processing to significantly reduce influence of these disturbances. Frequent registration and fast analysis of the video streams can assist fast necessary decision making and support planning of repair procedures, significantly lowering the maintenance costs and increasing the operational safety of underground infrastructure.