Object Recognition and Localization

Recognition and pose estimation result for a typical scene. The computed pose has been applied to the 3D object model and the wireframe model has been overlaid.
Recognition and pose estimation result for a typical scene. The computed pose has been applied to the 3D object model and the wireframe model has been overlaid.
Illustration of the performance of the developed features. The final filtered set of feature correspondences is visualized by the green lines.

Object recognition and 6-DoF pose estimation is one of the most important perceptive capabilities of humanoid robots. Accurate pose estimation of recognized objects is a prerequisite for object manipulation, grasp planning, and motion planning - as well as execution. Our research focusses on developing real-time methods for object recognition and in particular accurate pose estimation for these applications, using the stereo camera system of typical humanoid robot heads with a baseline comparable to human eye distance.

The listed publications deal with two different classes of objects: textured objects and single-colored objects. For the first class of objects, an approach based on local features is used, incorporating our developed high-speed features as a combination of the Harris corner detector and the SIFT descriptor. The developed approach for the second class of objects combines appearance-based view matching, stereo triangulation and 3D object model information. In both cases, special emphasis is put on accurate stereo-based 6-DoF pose estimation. With these systems we can capture 6-DoF object trajectories at frame rates of up to 23 Hz (textured objects) and up to 50 Hz (single-colored objects) using conventional hardware.