Vectorization of raster mechanical drawings on the base of ternary segmentation and soft computing

Abstract

The problem of automatic conversion of engineering drawings from paper carriers to an electronic vector form is very topical and is not solved yet at an acceptable level. The drawback inherent in many existing approaches lies in the fact that they are based on binary image segmentation. Under the conditions of low image quality, absolutely correct binarization is unattainable. Segmentation should be more flexible. It should divide pixels of the image not only into those belonging to the background and the objects but also take into account the existence of intermediate uncertain states. The purpose of this work is to increase quality of automatic vectorization of drawings having ambiguous situations: badly traced lines, areas of convergence and intersection of lines. The proposed approach includes stages of ternary segmentation of an image and fuzzy synthesis of a skeleton. Presented results of experiments show that, for grayscale and color drawing images of medium and low quality, the proposed approach provides better results than known methods.