Scene recognition with a monocular, accelerated moving camera in industrial robotics



The 3D-reconstruction of object depth information is very important for many technical applications such as 3D object recognition and classification, industrial quality inspection, driver assistance systems, robot navigation and satellite image analysis.


Within a DFG research project a measurement system for scene recognition is developed, which reconstructs the 3D object depth information with an accelerated moving smart camera and a 6-DOF inertial sensor.

The measurement system calculates the 3D object depth information from the camera pose, which is determined by the 6-DOF inertial sensor, and the pixel correspondences, which are detected on the basis of the optical flow of the image sequence. Due to acceleration of the camera additional information is incorporated in the reconstruction of 3D structures and this enables the 3D-reconstruction of a moving object with a monocular camera. The potential of this 3D reconstruction technique can be exploited in the production technology, where more and more industrial robots are used.

This research project is funded by the German Research Foundation DFG.