Semi-automatic analysis of rock fracture orientations from borehole wall images

We develop a semiautomatic method of identifying rock fractures and analyzing their orientations from digital images of borehole walls. This method is based on an algorithm related to the Hough transform which is modified to find sinusoidal rather than linear patterns. The algorithm uses the high-intensity contrast between the fracture aperture and the rock wall, as well as the sinusoidal trajectory defined by the intersection of the borehole and the fracture. The analysis rate of the algorithm itself is independent of fracture contrast and network complexity. The method has successfully identified fractures both in test cases containing several fractures in a noisy background and in real borehole images. The analysis rate was 0.3-1.2 minutes/m of input data, compared to an average of 12 minutes/m using an existing interactive method. An automatic version under development should open new possibilities for site characterization, such as real-time exploration and analysis of tunnel stability and support requirements as construction proceeds.