抄録
Normalized correlation template matching is applied to industrial image processing fields. However a target image area must be searched has the right potion and the same scale as a registered template image. This paper proposed a method to search a target image area that has free location and inclination in a gray-scaled image of printed circuit boards. Because electrical components and wires are set vertically and horizontally on a printed board, it is likely that edge directions are distributed vertically and horizontally a lot. The proposed method can estimate the inclination of a target image area by feature of edge directions and can reduce the computational cost for template matching. The inclination is measured by the edge direction histogram based on edge strength. The experimental results show that the inclinations of target image areas were measured with the averaged accurate of ± 2 degree and the computational cost was reduced to about 1% in comparison with the conventional method.
