Interactive Media Systems, TU Wien

3D object recognition and reconstruction from depth sensor data

Master Thesis, Bachelor Thesis, or Practicum

About This Topic

Experiment with the newest depth sensing technologies and develop tools to test their suitability for certain object recognition and/or reconstruction tasks.

Additional Information

The performance of different depth sensor technologies (time-of-flight, structured light, stereo and other recent technologies) depends to a large degree on the material and structures of captured scenes. To identify the properties of the different technologies, representative real-world scenes should be created and simple test cases implemented, so the sensors can be adequately compared. The test tools should be based on a software framework, which already integrates various depth sensors and provides methods for depth data processing such as the Point Cloud Library (PCL), to avoid integration/implementation overhead. Based on the chosen framework simple applications should be implemented to show the performance in standardized scenes. The effort of this topic can be adapted to different courses or theses, by choosing implementation of one or more test cases. For Praktika the topic is also suitable for a team of two.

Hardware

Intel Realsense, Kinect v2, Stereolabs ZED Mini, etc.

Software

flexibel z.B. Point Cloud Library, C++, Python