A Testing Method and Cognitive Model of Human Diagram Understanding for Automating Design Sketch Recognition

Sketches, whether hand-drawn or computer generated, are a natural and integral part of the design process. Despite this fact, modern day computational design tools are ill-equipped to take full advantage of sketching input. The computational challenges of recognizing sketches are easily overcome by human visual recognition and much insight stands to be gained by emulating human cognitive processes. Creating robust, automated tools that overcome the ambiguity of sketching input would allow for advances not only in the practice of engineering design, but in the education of design itself. One first step toward the development of a robust sketching tool is to determine how humans interpret mechanical engineering diagrams. This paper presents two contributions toward the goal of an automated diagram understanding system. First, a method is presented to gain insight into human diagram recognition using techniques analogous to peripheral vision and human attention. Following this, a cognitive model of human diagram understanding is presented from which to further develop computational design tools. With this work, researchers should be able to (1) improve understanding of human diagram recognition and (2) use our model to emulate human diagram recognition in future computational design tools.