Abstract
In industrial inspection, the accuracy of locating the edge points is very important. Sub-pixel is one of the methods that can improve the accuracy of edge location. Applying sub-pixel techniques to locate edge points on line features has been discussed for several decades, but few research discusses the feature of curve. Circle is an often used feature in workpieces and the location of edge points on a circle should be concerned. A new developed algorithm, which is named OCD (Orthogonal Circular Detector), is presented in this research. It consists of five 9x9 masks based on a set of truncated basis system and these masks represent a circular detecting area. In stead of using all the image data in the circular detecting area, only those on the periphery are utilized to save the computation time. When a segment of a circular object is located in OCD, two intersected edge points can be estimated by the OCD. The coordinates of the two edge points are not limited on integers such that sub-pixel resolution can be obtained. Experimental results show that significant improvement in accuracy can be achieved comparing to other existing methods. This achievement indicate its great potential in industrial application.