Modern approaches analysis of raster drawings digitalization automation

Nowadays, manufacturers, as well as distributors, have a large amount of paper-based technical drawings. The development of information and communication technologies last year cause the task of such drawing digitalization. The task becomes popular 20 years ago in the research and development community (R&D) and now it is still opened since it has not been solved completely. Now there is a lot of scientific papers as well as a lot of commercial systems that have been created to support such a process. R&D community considers the following tasks: converting a raster drawing to vector, noise removing, etc. Some of them can recognize scanned drawings with good quality. The paper aims to analyze current research and developments for the raster drawings digitalization automation. We consider and analyze different approaches that allow supporting the digitalization process: neutral networks (YOLO, MFC-gan, AUTOFEAT, intelligent feature recognition methodology), graph theory-based approach, an algorithm based on a set of rules and Opitz coding system, an algorithm based on field losses of the surface and the concept of bridges. There is commercial software that supports the digitalization process for the PDF that has been exported from the CAD software. However, scanned image recognition does not work fine.