Robust optical measurement of dust thickness on a flexible filter surface

Abstract

A stereo computer vision system as a thickness measurement device is introduced. The challenging task is to measure the thickness of dust deposits over the surface of a flexible cloth filter with a target resolution of 0.5 mm2 per measurement in the filter plane and a depth resolution of up to 50 μm. The total area to be measured is 100 mm × 2500 mm. There is practically no texture present on the dusty filter surface. Analyzing dust deposition over the whole filter length during operation has not been possible so far and promises to give insight to air flow in the filter chamber, behaviour of different filter materials and design flaws of filter chambers. Our approach is to first create dense 3D-reconstructions of surface patches. This is done using a calibrated stereo setup and a standard slide projector that provides texture by projecting a random pattern onto the filter surface. The thus obtained surface patches are registered against each other using an ICP algorithm to get a complete 3D model of the filter surface. This step is done twice, for the clean filter surface and for the surface including dust deposits. We finally retrieve the actual depth information by registering the clean and dusty models using markers applied to the filter surface and calculating the difference. In this paper we present the measurement device as well as the underlying vision system and show that it is possible to meet the given requirements under industrial conditions.