Advanced Quality Inspection through Physics-Based Vision

Physics-Based Vision is a combination of traditional machine vision, material optics and computer graphics. Illumination, optical reflection and image formation are analysed in this new field of research in order to develop more efficient artificial vision systems. This paper focusses on two aspects of physics-based vision: analysis of spatial distribution of reflection as well as polarisation by optical reflection. Highlights are a common phenomenon in images of real surfaces, that may cause standard vision algorithm to produce erroneous results. Algorithms that can eliminate these highlights are introduced in this paper. On the other hand, analysis of specular reflection, the physical effect that causes highlights, provides valuable information about the inspected surface. A single unpolarised ray of light that is reflected on a surface will cause a complex distribution of light intensity around the ideal angle of reflection. In addition, the reflected light will be partly polarised. Surface properties such as gloss and roughness can be described by these physical effects. Inspection of printing quality of cigarette packs is an example of a machine vision task that can not be solved by traditional methods.