Non-contact surface inspection
This thesis describes the development of a non-contact inspection system, used in providing a comparative method for gauging a surface. The purpose of this system is not to measure the 3-D shape of a part. The volume difference between the part to be inspected and a master, is merely part of the decision criteria, where if this difference exceeds a certain threshold value, then the component under inspection is deemed to have failed. The technique involves a combination of fringe projection and image subtraction. The system comprises two sub-systems, a low-cost pc based image processor and a white light, square wave fringe projector. A camera provides the interface between the two sub-systems. Validation of the technique is provided by the simulation of mathematically generated defects, and by means of experiments on samples of known volume. In addition, the effect of the variation of particular set-up constants, on the technique's accuracy are also illustrated. The problems and subsequent solutions associated with the practical inspection, result in an improved method of gauging. The system provides reliable results (within 4%) for surfaces of nominally similar form and reflectivity. Additional results (within 10%) are illustrated, where the images and fringe patterns are mis-aligned. As a result of a selective filtering scheme, precise relocation of the surfaces (used in the comparison) is unnecessary. This is conditional upon the fringes on each surface, being identical in orientation with each other. Further consideration of the technique (within an error analysis) indicates the necessity of an accurate determination of all the set-up constants. The error results cannot be taken too literally, since the worst case values are presented. However, the trends of the effects of these errors are useful. An adapted alternative method is also described that may prove to be (in certain applications) an interesting real-time solution.