Using Image Processing for Architecture Extraction from Non-Standard Sources

Many organizations, programs and teams have yet to adopt or implement modern model-based systems engineering tools, and instead continue to represent their systems architectures using virtual "paper-based" formats such as PDFs or images. A system's architecture is a crucial portion of the technical baseline that engineers and decision-makers alike will continuously develop, maintain, and reason over during the lifetime of any system. This paper presents a new tool, ArchEx, that lowers the barrier-to-entry for those looking to transition beyond these static architecture images. Using a combination of Python-based computer vision techniques, ArchEx allows for the intelligent, automated import of virtual paper-based architecture artifacts into commercial tools to allow for digital model representation and manipulation. It parses the architecture image to derive a graph structure which is then loaded into a model-based systems engineering tool. ArchEx provides an automated assist to the systems engineer, decreasing the time and effort typically required to transcribe image-based documentation into a modeling tool. It is being actively developed within the MITRE Corporation, with initial test applications and experimentation showing promising results. This paper will provide an overview of the technical approach surrounding ArchEx, a visual demonstration of the capability in its current state, and a brief summary of planned next steps, and future applications.