FAULT DETECTION OF PROJECTION ON CURVED SURFACE

Abstract: In this thesis a system that evaluates projection systems is designed and tested. The type of projection systems of concern feature display surfaces that are curved in two axes. The system uses a calibration image featuring evenly spaced circles. This image is projected onto the display surface. Photos of the projection, captured using a camera placed in the design eye position of the projection, are stitched into an equirectangular projection of the display surface. The resulting system identifies both absolute and relative pixel position errors of the projection by comparing the location of the projected circles with their intended location. The developed system can detect faults in systems projecting on curved surfaces. There are several sources of error in the resulting system. The panoramic stitching and is a minor source of error. The largest source of error is the Hough transform used to detect and locate circles in the projection. This transform is not very robust. In pixel perfect images the system returns perfect measurements, however, if images are captured by a camera, the Hough transform lacks accuracy. The largest measured measurement error for the absolute pixel position error is approximately 10.6 arcminutes. For the relative pixel position error the largest measured measurement error is approximately 0.85 arcminutes per degree. The developed system can be used to acceptance test projection systems as well as indicate needs for maintenance of them.