Light field cameras and reconstruction of specular surfaces

The increasing demand for specular free-form surfaces has posed a major challenge for production and measurement technology in recent years. When inspecting reflective components, such as solar collectors, lens systems or telescope mirrors, one is generally interested in the most precise manufacturing possible and thus one is interested in the exact geometric proportions of the object. The reflective property also plays a decisive role for surfaces used in the automotive industry, e.g. painted chassis parts, because here faults and defects in the paint greatly influence the aesthetics of the product.

The inspection of such surfaces is very demanding in practice. In contrast to diffuse reflection, an observer does not see the specular surface itself but the distorted mirror image of the environment. The reflective surface is virtually invisible to the observer. Automatic visual inspection, especially 3D reconstruction, is therefore a challenging task.

The deflectometric measurement techniques use the law of reflection and the knowledge of the arrangement between a camera and a pattern generator to draw conclusions about the shape of the surface from the deformations of the mirror images. For automatic visual inspection and exact 3D reconstruction, precise knowledge of the system parameters is required (e.g. size and position of the LCD monitor relative to the camera sensor, as well as the intrinsic camera parameters). Even with complete knowledge of the system, a single camera is generally not sufficient to calculate a unique surface from the reflection data. Possible approaches are stereo methods, optical flow or model-based approaches.

Research will concentrate on the development of novel methods that enable high-precision measurement of specular surfaces.