Interest of Syntactic Knowledge for On-Line Flowchart Recognition

Abstract

Abstract-In this paper, we address the problem of segmentation and recognition of on-line flowcharts. Flowcharts are bi-dimensional documents, in the sense that the order of writing is not defined. Some statistical approaches have been proposed in the literature to label and segment the flowcharts. However, as they are very well structured documents, we propose to introduce some structural and syntactic knowledge on flowcharts to improve their recognition. For this purpose, we have used an existing grammatical off-line method with on-line a posteriori signal. We apply this work on a freely available database. The results demonstrate the interest of structural knowledge on the context to improve the recognition.

Online flowchart understanding by combining max-margin Markov random field with grammatical analysis

“…al. [18] analyzed the structure of the flowchart and described them in DMOS (introduced in section 4.1), a grammatical method for structured document analysis. This approach provides a global solution to the problem.…”

“…This second step does ensure the global consistence of the recognition. The presented system is an extension of the grammatical analysis from the previous works [18] and [8]. In this former solutions, the analysis was based only the set of atomic elements in data: the full strokes and the straight-lines (segments) within the strokes.…”

“…Using these terminals, their relative layout, and the a priori lexical knowledge about flowcharts the syntactical analysis recognizes the symbols. Compared to [18] this new work adds two main propositions: using pre-recognized strokes from a statistical approach (the MRF in our case) and producing the relationship labels between the recognized symbols.…”

“…The parameters of M3N are learned using structured SVM. After stroke labeling, we use a grammatical description language extended from a previous work [18] to parse the structure of a given sketched diagram, thus obtaining the final recognized flowchart. In order to fully evaluate of the recognition process, the ground-truth of the flowchart dataset [2] have been completed by adding the object relations.…”

Logical Structure Recognition of Diagram Images

“…In [4], the analysis to label each stroke of the flowchart and to group the strokes depending on the symbol they belong to is presented. The same area of research is represented in the paper [5].…”