A fiber-optic illumination device that aids the inspection of window surfaces has been improved to make it suitable for automated detection of pits, scratches, and subsurface damage by use of a machine-vision and image-analysis system. The improvements, which include modifications in both the optical and mechanical aspects of its design, significantly reduce the brightness of stray light that, previously, was bright enough to interfere with automated detection of defects.
The previous version of the device (see Figure 1) was described in "Illumination Device for Inspecting Window Surfaces," NASA Tech Briefs, Vol. 22, No. 11 (November 1998), page 68. The device could be temporarily attached to a window by use of suction cups. The device coupled light from a fiber-optic cable, through a block of clear material, into the window pane to create a side-lighting effect. One advantage of this or a similar lighting scheme is that light remains within the window because of total internal reflection, except at sites of surface or subsurface defects. Hence, the defects appear bright on an otherwise dark window surface and could be readily observed. Another advantage of this lighting scheme is its selectivity: The more important deep defects present larger cross sections to side illumination than do shallow defects of the same lateral dimensions and therefore tend to appear brighter. Furthermore, dirt particles on the surface, which do not constitute defects in the window and are thus unimportant for purposes of this type of inspection, may couple some light out of the window but they also shield the observer from this light. Hence dirt tends to appear dim, and the incidence of false detections of defects is lower than it would be if a different lighting scheme were used.
A human inspector can distinguish between foreground and background light by, among other actions, refocusing at different depths and changing the angle of observation; however, an automated defect-detection system cannot do this. Therefore, to ensure high sensitivity and consistent performance in an automated defect-detection system, it is necessary to minimize background light, including stray light generated by the illumination device.
The previous version of the illumination device generated appreciable stray light because of two weaknesses in its design. The first weakness was that a significant fraction of light was not subject to total internal reflection in the window. The second weakness was that a significant amount of light was scattered by partial reflection at (1) the interface between the acrylic block and the window and (2) the end of the fiber-optic light guide and the adjacent poor-optical-quality surface at the bottom of the light-guide hole in the acrylic block.
The improved design overcomes these weaknesses. The acrylic block of the previous design is replaced with a triangular prism (see Figure 2) that is shaped and dimensioned to minimize undesired scattering and maximize the amount of light subject to total internal reflection. Light from the fiber-optic light guide enters the prism through the smallest prism face, which is nearly perpendicular to the window surface. Unlike the bottom of the hole in the acrylic block of the previous version, this prism face is accessible prior to assembly, and can therefore be polished to minimize scattering.
The angle of the fiber-optic light guide is adjusted so that even the light rays that enter the window pane most steeply (e.g., ray A in Figure 2) are subject to total internal reflection. Because of the large angular spread of light leaving the fiber-optic light guide, some rays (e.g., ray C in Figure 2) go toward the surface of the prism, where they are reflected downward toward the window pane. Proper choice of the prism angles ensures total internal reflection both at the top surface of the prism and the bottom surface of the window pane.
Because of a small mismatch of the indices of refraction of the prism and window, a small fraction of rays is reflected at the prism/window interface (e.g., ray B in Figure 2). In the previous version, these rays contributed to stray light. In the present version, these rays are reflected from the top surface of the prism back toward the window.
The prism must be made short enough to prevent coupling out of rays that have been reflected once from the bottom surface of the window. The maximum length permitted by this criterion depends on the thickness and the index of refraction of the window.
Going beyond purely optical considerations, another disadvantage of the previous version of the device was that as the suction cups gradually relaxed, a gap of approximately 1 mm developed between the window surface and either the acrylic block or (if used) the clear rubber sheet attached to the acrylic block. It was necessary to fill the gap with an index-of-refraction-matching liquid, which could easily run out or evaporate under the heat of the illumination, making the device unreliable after a short time.
In the improved version, the prism is mounted in a frame with a free range of motion, >1 mm, that makes it possible for the prism to remain in contact with the window. After attaching the device to the window by use of the suction cups, one pushes specially shaped wedge slides between parts of the prism and the frame. This action stretches the suction cups, providing sufficient force to ensure steady contact, thereby retarding the loss of coupling fluid. The net result is that the device provides reliable and constant lighting with minimum background illumination suitable for automated inspection of windows.
This work was done by Henry Weidner, Terry Greenfield, and Carl Hallberg of Dynacs Engineering Co. and Anthony Kraljic of United Space Alliance for Kennedy Space Center.