Software development for the recognition of the real volume and convex hulls of objects by 3D scanners. Sensor technology and 2D/3D graphics for measurement, object recognition and calibration of laser. Development of algorithms for above mentioned areas. In detail: rotating calibers, minimal bounding box, ransac, eigenvalue calculation and others. Pre-calculation and testing of algorithms with Matlab.
Customer: International sensor manufacturer.
Techniques: C++, Visual Studio, Matlab, OpenCV, PCL, Halcon, Eigen, UML, SVN, Blender, AutoCAD
- read more
For an international sensor manufacturer, we developed mathematical models and software algorithms for the evaluation of 3D scan data. For the evaluation, numerous filters were applied upstream to improve the calculations and optimize the output data.
The task was to recognize and classify objects from the 3D scan data (point cloud), i.e. the recognition of the geometric shape (cylinder, cuboid, sphere, etc.). A subsequent calculation of the volume brought the desired results with fast performance.
In the test phase, we installed the scanners at various positions in the laboratory, connected them to the embedded system and configured them. Data from a production plant was also used, to which the newly developed 3D scanners were pre-assembled. In order to produce an abundance of variants for the tests, we modeled a setup in 3D for a scan data simulation.
kununu