Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB

3D Analysis

3D analysis is commonly understood as the acquisition, processing and evaluation of 3D data. The data is available in the form of point clouds. These can be obtained directly by means of special 3D sensors, for example laser / lidar sensors.

However, the focus of the VID department is on the image-based acquisition of 3D information. Three basic methods are used for this purpose:

  • Multi-dimensional 3D reconstruction
  • Motion-based 3D reconstruction
  • Model-based 3D reconstruction

As with the human eyes, multi-dimensional 3D reconstruction is based on the fact that several cameras look at the same scene from different angles.

In motion-based reconstruction, the same effect is achieved with only one camera by calculating the displacement of the camera between two points in time.

Model-based methods require a more or less exact 3D model of an object or a scene. This model is projected into the image and fitted into the existing structures. In this way, perspectively represented objects can be metrically recorded.

3D reconstruction from a monocular aerial sequence

Generated polygon model from an unstructured set of 3D points

Adaptation of a road and vehicle model for the three-dimensional detection of the scene

Obtained depth map from a stereo camera pair