A new approach to a Semi-automatic discontinuity sets extraction from point clouds .

Sensor such as light detection and ranging (LiDAR) or SfM (Structure from Motion), point clouds (PC) are nowadays an essential tool for the rock instabilities analysis. PCs currently allow us to images complex 3-dimensional discontinuities. Thanks to the high density of data and the high accuracy of the LiDAR have the potential of a semi-automatic fault identification. Previous work has already tackled the question by using methods such as least square analysis or the normal vector orientation calculation and other more complex method. Those methods where successful but the accuracy on fault detection were not sufficiently high enough. In order to overcome those encountered issue, the development of a new kind of fault detection algorithm were needed. During our work we have developed a new semi-automatic method of fault identification using the variation of point density. The developed method has successfully detected discontinuity as well as their orientation and their number. The 3-dimentional scanning of the PC by the algorithm allow us to have a good redundancy even on complex fault shape. Results on simple synthetic data are convincing enough to test our algorithm to more complex synthetic data with more randomized structures. Some test can be also be done on simple LiDAR dataset(simple shape (cube)) to consider instrumental noise or potential artefact before experimenting the algorithm on more real data. In the future this work could lead to data analysis of the output of the algorithm in to determine the frequency of similar discontinuity that can lead to the estimation of the potential volume of material that could be in movement.