Complicated Fracture Detection And Evaluation Based On Curvilinear Feature And Geometry Analysis

The fracture detection and evaluation are significant for engineering project assessment and geological hazard prediction;in practical applications,the development degree of fractures is usually used for judging whether the engineering project can be implemented or not,predicting whether the geological hazard will occur or not.Conventional method for fracture detection is the site measurement by manual,which is not only labor and time consuming but also dangerous;fracture evaluation relies on human experience,which is not objective.Therefore,a convient and reliable method for fracture detection and evaluation is necessary to be proposed.In this paper,according to the characteristics of fracture image,the key techniques,which contain template extraction,centerline extraction and edge detection,feature extraction and comprehensive evaluation,are discussed in detail.The concrete research contents are as follows:1.Fracture classification and pre-processing.The characteristics of different types of fractures are not the same,which leads to various methods for processing and analysis;so it is necessary to discuss their similarities and differences.Most of images acquired by digital camera are color level while the following algorithms are mainly suitable for grey level images;therefore,color transformation demands to be implemented.In the process of image acquisition,due to the interferences by digital camera and external factors or the rough surfaces,there are many kinds of noises on fracture image;these noises should be filtered out.2.Fracture template extraction.Fracture template can be used for accelerating the following centerline extraction and edge detection.The paper makes use of its curvilinear property to process the pre-processed fracture image.Firstly,the Frangi filter is adopted for multiscale enhancement of curvilinear structures;then,conservative threshold,which is selected based on histogram analysis,is conducted for image binarization;finally,the fractures are screened and extended by using object analysis and mathematical morphology processing.3.Fracture centerline extraction and edge detection.Fracture centerlines and edges are both located on the pre-processed fracture image in range of template,making use of its curvilinear property.For centerline extraction,the Steger algorithm is firstly used to detect and connect the centerline points and then the centerline segments are linked based on distance and angel difference.For edge detection,it is mainly based on the improved Live-wire Contour algorithm.Firstly,Canny detector is used for constructing the cost function;secondly,seed edge points are searched in the direction perpendicular to centerline;finally,these seed points are linked in order by optimal path searching.4.Fracture feature extraction and comprehensive evaluation:After fracture location,the geometry features(the ratio of fractures,the average number of key points and length per fracture,fractal dimensions of fractures related to roughness and distribution respectively)are statistically estimated;and then,according to fuzzy comprehensive evaluation model,some of the features are chosen to estimate the evaluation indicator,which can be used for describing the complexity of fracture surfaces.