Research Of Tunnel Face Geology Information Based On Image Processing

Research of tunnel face geology information on image processing analyses tunnel face images by modern computer image processing technology to grasp geological condition of the excavated tunnel faces and their frontage . So that provides important information for tunnel design and construction. A lot of works have been done on tunnel face images on abroad, and some tunnel face image processing systems have been used in projects. Domestic research is started lately, so that the developed systems which are not mature are hard to apply in practical.This thesis makes research of tunnel face images centering on image processing technology. We study a method of automatically extracting boundary lines combining with typical image processing algorithms.The method uses Canny operators having high accuracy positioning and good noise-immune capability to detect edges. Then single-pixel wide borders are obtained by using mathematical morphology algorithm to further process the boundaries. In the processing of extracting boundaries, the paper studies single-point and node point elimination algorithm, boundary auto-connection algorithm and auto-filter algorithm. They realize noise decrease and auto connection with discontinuous boundaries and further process of noise after connecting boundary lines. After that we extract characteristic parameters, such as rock quality designation, crack length of unit area, average distance between structure planes etc, to realize automatical estimation about tunnel face rock. And then we identify the rocks employing neural network. Author uses texture descriptors and SFM (Spatial Frequency Measurement) ,when analyzing image characteristic of rock.The thesis applies these studied algorithms to actual environment. The results show that boundary auto-connection algorithm could completely extract boundary lines of structural planes, and generate geologic sketch map by adding a few manual operation that obviously shorts the time of sketching geology. And the extracted characteristic parameters of structural planes realize automatic analysis of tunnel face rocks. Moreover, the trained neural network can identify lithology accurately.