Identification of discontinuities in a rock mass from a three-dimensional point cloud using an SVM

Matsukawa, Shun*; Itakura, Kenichi*; Hayano, Akira  ; Matsuura, Yuto*; Suzuki, Yukinori*

A study on identification of discontinuities from estimated surfaces in a rock mass using an SVM was carried out. We extracted four features to identify discontinuities from the estimated surface: fractal dimension (FD), arithmetic mean roughness (Ra), maximum height of roughness (Rz), and roughness-length (RIM). The SVM was trained using combinations of these features. The performance of the SVM was an average of 0.274 in the F value.