Data reduction in 3D open cast models
| Runtime: | 31.10.2013 - 31.03.2015 | |
| Funding: | BMWi funded within the framework of the AiF-program ZIM | |
| Project staff: | Dr.-Ing. Matthias Klaus (former staff member) |
Motivation
In recent years, opencast mine operators have increasingly used systems for the holistic consideration of mining production processes. Essential functions are the permanent collection and consolidation of relevant operating data and the prompt provision of current site and equipment information in processed form for a variety of purposes such as performance calculations, safety checks or maintenance planning. Servers centrally manage the data supplied by many stations, generate the visualisation data and make it available on demand from clients. Problems currently exist here, as wireless networks and mobile devices are increasingly being used on the client side. A lack of bandwidth or excessive data volumes currently limit the application possibilities of the monitoring systems.
Objective
Fig. 1: Surface opencast mine (example)
The 3D information of the open pit surface originates from various data sources (overflight measurements, scanners on working equipment, calculations from geometry and movement of the cutting tools).
The example shown in Fig. 1 covers an area of approx. 5 km x 5 km and has elevation differences of ~140 m. The resolution in the plane is ~30 cm, Fig. 2. The amount of raw data is ~114 million points and ~228 million triangles.
Fig. 2: Excerpt of the open cast mine (example)
For data reduction, a method has been developed that works exclusively in the 2D image area. For this purpose, the 3D model is converted segment by segment into height images, Fig. 3.
Fig. 3: Conversion of the base data 3D-> 2D
The aim is to use only points of high curvature to generate the reduced 3D model. Areas in between are covered by triangulating the points.
Project content
Fig. 4: Data reduction level 1
The procedure itself works in two stages. First, salient points are extracted for each opencast mine segment, Fig. 4. The description of the opencast mine surface exclusively by these points usually leads to a very inaccurate reduced model. Therefore, in a second step, further surface points are specifically added to the reduced model, Fig. 5.
Fig. 5: Data reduction level 2
Result
Reductions to ~2% of the original data volume could be achieved by the procedure. The accepted error in the height values is ±10 cm.