Fracture characterization based on data fusion technology and its application in rockfall hazard assessment

Rockfall has become one of the deadliest geohazards in Southwest China and how to comprehensively and effectively assess rockfall hazards is an urgent challenge to overcome. Additionally, comprehensive characterization of fractures on rock mass outcrops is a prerequisite for detecting potential rockfall. In this paper, an image and point cloud-based data fusion technique is applied to characterize regional rock mass fractures. Firstly, the performances of three classical computer vision algorithms are compared and SegFormer is selected as the appropriate base model for fracture detection. After that, according to the coordinate projection transformation criterion, the detected fractures are mapped to the point cloud. The parameter information obtained through fracture characterization is used to develop a representative three-dimensional discrete fracture network (3D-DFN) and then according to the results of the volume distribution of rock blocks, the three frequencies (high-frequency, medium-frequency, and low-frequency) of rockfall events are numerically simulated to obtain the characteristic information of rockfall trajectories. Finally, based on the characteristic information of rockfall trajectories and the GIS platform, the risk of rockfall hazards with three frequencies is evaluated and analyzed. This paper provides a new way for geologists to assess the risk of rockfall hazards and propose reasonable rockfall hazard prevention schemes.