We developed an automatic matching algorithm, which detects the characteristic point based on the difference of color and texture, from the tunnel wall image captured with a plural line sensor camera, and thus we can obtain whole cross section images efficiently. First, we describe the developed sub-pixel style image-processing algorithm to extract minute crack from a standard wall image and measure the characteristic values, such as width, length, direction etc. at high accuracy. Furthermore, we developed the improved camera, lightning equipment and an additional function of focus adjustment to obtain a high accurate cross section image. As a result, the total cost has restrained and the practicability of the detection system has fairly progressed.

Copyright (c) 2006 Railway Technical Research Institute