We are developing an automatic image processing algorithm to extract deformations like cracks and diagnose tunnel conditions. As the first step for this purpose, it is indispensable to register plural images accurately and effectively from continuously scanned images of tunnel wall. Because there are problems in the accuracy and work efficiency in the manual image registering process to appropriately align corresponding points, a method is proposed to automatically perform this process. This method extracts cracks on standard tunnel walls. We applied various techniques to this algorithm. One is to distinguish cracks in the obtained images of tunnel wall. As there are several noise elements in those images however, we must separate cracks from noises. The other is to measure various dimensions such as the length or width of cracks. We can prognosticate the growth of cracks, and rank the degrees of soundness based on these data.

Copyright(c) 2001 Railway Technical Research Institute