Combining global and local vision for technical document understanding

We have implemented a method that describes as precisely as possible all the shapes of an original technical drawing during each level of the analysis. We integrate the natural techniques that normally associated with human perception, and our new representation model provides both essential information about the image and ways to manage this information. It enables a global vision on which the study of the document can be based. Some local analyses confirm, invalidate or improve this information and make the context and the representation evolve in order to obtain data of a higher semantic level