Remote sensing and mapping of geological rock-mass features employing advanced data analytics and artificial intelligence

Abstract

The mapping and classification of geological rock mass features like discontinuities, block sizes, etc. are important information for the profound understanding and prediction of the rock mass behavior under various load conditions. These are the basis for stability assessment of rock slopes, the evaluation of geological hazards, and the design of mitigation and support measures. Manual data acquisition is often time-consuming and possesses a certain risk to human life, especially in remote areas after hazardous events. Another challenge is the mapping of large and difficult to access areas. A team of AFRY Geologists, Geotechnical Engineers and Scientists developed a novel software to map and characterize geological rock mass features based on point cloud models. The latter can be developed from various remote sensing techniques. The software uses a self-supervised neural network that provides k-d tree 3d directional search from point cloud neighbors. On the basis of the neighbor normal estimation, the best plane fit is evaluated. For each point pair, the dip angle and dip direction are calculated. The same approach can be used for the characterization of any other feature. The two characteristics/values (dip direction and dip angle) are employed to e.g. classify discontinuities in groups based on pre-specified angular ranges. This paper presents the introduced remote mapping process and results with the help of two case studies. Furthermore, insights concerning the data analysis process and evaluation approaches are provided.