Hochschule Ravensburg-Weingarten | Zurück zur Startseite


Nov. 2016 - March 2019, on behalf of FHProfUnt


Schematic representation of the fusion between a PMD-picture and a 2D color picture

The joint research project Smart3D with the IFM group is developing innovative AI-based sensor fusion and object classification by harnessing the fact of having a huge amount of data values, whose patterns are examined and recognized via deep-learning networks. Sensor fusion is about improving a low-resolution PMD picture by using the information of a high-resolution 2D colored picture and matching the color information with the corresponding information of the PMD picture on a pixel-based level.

Unfortunately, the use of simple interpolation methods is not successful, especially regarding the transition of objects in combination with a far-away background. A new research approach is based on machine learning methods by training a neural network with different data sources. This approach is transcribed to PMD data within the scope of Smart3D.

If interested or in case of any questions:
Markus Schneider