3-D Inspection System for Small Defects on Arbitrarily Curved Surfaces

Abstract

This paper describes a three-dimensional (3-D) inspection system for small 3-D and reflectance defects of an arbitrary surface. We fix a TV camera in front of the object, and along the viewing axis we move a random dot projector at a constant speed. After taking the first image sequence, we shift the object by a slight known amount, and take the second image sequence in the same way. Applying the spatio-temporal gradient method, we extract gradient patterns of surface depth and surface reflectance distributions. We applied this method to visual inspection and classification of fine surface defects such as cracks, flaws, stain, and rust. We show this system could detect up to 5μm flaws and 1% reflectance change from a 50cm distant measuring apparatus.