An auto-focusing microscope for surface inspection


Student thesis: Master's Thesis

View graph of relations


  • Chiu Chun NGAN


Awarding Institution
Award date4 Oct 2004


This project studies the development of an automated imagining system for the inspection of large curved surfaces. Positioning of the imaging camera is supposedly under numerical control. Without loss of generality, the motion platform is based on a 6-axis PUMA robot. The focus is on imaging applications that require large magnification. The field of view is usually small compared to the combined positioning error due to the repeatability of the robot and the tolerance and positioning of the surface for inspection. This problem of insufficient positioning accuracy of the primary motion system is alleviated with the introduction of a motion sub-system at the camera mounting. In the current implementation, a linear slider is mounted at the end-effector of the robot to enable fine motion of the camera along the normal direction of the surface. The strategy is to take a series of images of the surface as the camera travels in-and-out of the focusing range. It is known in Fourier optics that a well focused image has higher spatial frequency content than one that is out-of-focus. Based on this, a fast and simple algorithm was implemented for selecting the best focused .image. The algorithm involves computing the standard deviation of grey scale values for all pixels of an image. Two imaging functions have been implemented: (i) capturing an image at a specified location of the curved surface; and (ii) capturing images along a specified path on the curved surface while the robot is tracking the path. These functions, particularly the second, requires coordinated motion of the robot and the linear slider. An application of part surface polishing inspection has also been studied which is suitable for the proposed imaging system. Polishing is a multi-stage process. Grooves are developed on the surface during polishing and their direction follow the direction of the rubbing of abrasive particles. Suppose polishing is planned such that the direction of rubbing is distinct for each location of the surface between the current stage and the previous stage of polishing. The change of direction of the grooves may then be used to flag the completion of the current polishing stage. A non-contact technique for the on-site inspection of moulds and dies polishing is explored. The technique is based on analyzing the directions of the grooves from images of the surface being polished. Experiments have been conducted to validate the proposed inspection technique when the surface roughness of a part surface was reduced from 0.9pm to 0.02pm during polishing. The results are promising and show that the proposed approach could be practical and robust.

    Research areas

  • Microscopes, Surface roughness, Surfaces (Technology), Measurement