Crack edge detection of coal CT images based on LS-SVM

A uniaxial compression scan test for coal specimen with the industrial computer tomography (ICT) equipment and the loading system were carried out. Clear CT images were obtained which indicate development of microcracks within the coal specimen at different stress stage. The CT image intensity of neighborhood of every pixel is well estimated by least squares support vector machines (LS-SVM) and the gradient operators and zero crossings operators are obtained. The edge detection method based on combination result of gradient and zero crossings to acquire the crack and the hole edge is proposed. Analysis the same scanning section´s four differ stress stage´s CT images and withdrawn the image crack and hole part. Clear crack and hole region space distribution has obtained. The crack or the hole region size distribution changed along with the differ stress process. Compared with Canny algorithm, experiment results of coal CT image crack edge detection by LS-SVM are satisfied.