Rock Mass AI

Strayos features a powerful Rock Mass AI module, where it will internally calculate and cluster groupings of Geological features within a rock face profile to show Dip and Strike directions and angles. To begin, navigate to the site you are looking to address. Select the Rock Mass AI module, where it will then bring you to a 3D view of your model.

From here, users can toggle on the clusters which will showcase the lines and planes of each cluster of features detected by the AI. These features can then be downloaded and exported in the following formats: 

Additionally, a Stereonet can be viewed from the AI detected data showcasing the geological information derived: 

To further read upon the algorithms that produce this data, see below: 

1. At the first stage, Strayos Rock Mass AI module will detect the crack polygons on the surface of the bench, it can’t be able to distinguish between joints and other random cracks.

2. A the second stage, we designed a clusterization algorithm to cluster those cracks in the several groups. It basically tries to fit the cracks on planes that have the same orientation. If majority of the cracks can be fitted to such planes, they will be clusterer into the same group. This algorithm currently has two limitations, first one is that it requires the majority of the detected cracks are formed by the joints. 2^nd is that all the cracks can be easily fit to the bench surface direction.

For example, in the image 1 below, we can clearly see three different joints group. But since all the cracks are on the bench surface plane, it means that all the cracks can be fitted into this group(image 2). Therefore, we assign less weight to such group, which means if there are other good groups that can be fitted with majority of the cracks, even they have lesser cracks, they will have higher possibility in the joints group in the final results(image 3). But as I mentioned, this still relies on that majority of the cracks are formed by joints. In the images from above analysis, you can easily observe such groups on bench surface.

Our approach based on predicting a crack-polygons on the terrain's surface and fitting planes through that polygons on the terrain.

Our planes describe the inner structures like crack/bedding planes

For example “Cluster 4” (image below) summarizes planes which points to a big sloped crack. It is not a sort of outer surface structure. It is plane surface which describes the inner structure.

Here are steps of calculations:

•        Predict joints on the 3D textures.

•        Map predicted 2D-polygons to 3D-model and obtain 3D-polygons

•        Fits 3D-polygons by planes

•        Found Strike and Dip angle for each predicted plane.

◦     "Strike" is the azimuth of an imagined line which obtained as intersection fitted plane and horizontal plane.

◦     "Dip angle" is the angle of inclination measured downward from horizontal.

◦     Sometime useful operate with "Dip direction", which is projection of downward to horizontal plane and equal to Strike + 90[deg].

•        These data related to lines and described on the stereo net by red-points as plane's poles.

◦     Example how to put a pole into stereonet: https://www.youtube.com/watch?v=8JOzmIBVjkk

•        Then data is clustered and obtained clusters are approximated. Approximated clusters in stereomap named as planes.

If point the camera view from the top to the reconstructed model and use button “Align view with North” it will match vertical axis with North. Slope of crack-plane described by dip-angle 57 [deg]