UML Sketch Recognizer: um aplicativo para reconhecimento de esboços de diagramas de classe em fotos

Abstract

The Agile Modeling is a software modeling strategy for agile development. There are several practical to perform a responsive modeling, such as the use of simple tools. The whiteboards are used by many teams that use agile modeling practices, they offer a dynamic and collaborative environment modeling. The problem occurs in the generation of software artifacts. The sketches on whiteboards are usually stored in photos, causing a downgrade of the documentation and also inability to generate the outline equivalent codes. In order to minimize this integration between sketches in frames and UML editors, this paper presents a recognition application class diagrams of sketches for Android. The application uses processing techniques and image analysis with computer vision library Open CV and interpreter Tesseract characters. The interpretation of the image is made by a recognition layer processing, segments, classifies and interprets the image containing the diagram. From tests with various sketches of diagrams, it was possible to realize a high level of detection of the elements in a controlled environment, which was around 70% and 100%. Already recognition had an average level of precision that was around 30% and 90%.