Sub-pixel edge estimation based on matching template

Abstract

The significant disadvantage of traditional sub-pixel edge location algorithm is difficult to achieve high accurate edge location of micro-parts, because reflective characteristic of edge is widely divergent for material, chamfer, and any other factors. This paper introduces a sub-pixel location algorithm for micro-parts edge which is based on matching template. First, it studies the edge imaging characteristic for a specific practical engineering part, and constructs one dimension matching operator. Second, we abstract the pixel edge points on the image of practical engineering parts, and get the normal direction information of the whole points. Finally, according to the correlation operation between matching operator and a series of pixel points by the normal direction of the edge points, found the best position for matching. This position is the location point of the edge point under sub-pixel accuracy. Compared with traditional sub-pixel algorithm, this algorithm requires different matching operators for different measured objects, not a universal operator to all types of the edges. Therefore, the effect which brings by the parts' edge characteristic and the imaging characteristic can be reduced. This algorithm has prominence significance for using sub-pixel technology to measure practical parts. Experiment result shows that, for the practical engineering part which has clear measurement aim and conformable character of the edge, this method will achieve high accurate edge location, and it is suitable for wholesale measurement of edge for complex micro-parts which require high accurate.