Knowledge-based recognition of crosshatched areas in engineering drawings

Abstract

One of the main tasks in the problem of engineering drawing (ED) automatic input and interpretation is recognition of ED primitives to obtain 2D image representation for further use in CAD systems. This paper presents two-stage knowledge-based algorithm to recognize crosshatched areas in ED images. At the first stage, pure segment chains i.e. simple structure crosshatched area parts satisfying a template description are detected. At the second stage, the final assembly of the area takes place by using the chains assembled at the first stage.