Intelligent Extraction of Multi-style and Multi-template Title Block Information Based on Fuzzy Matching

Abstract

ABSTRACT Title blocks are crucial primary data on drawings necessary for the operation of modern manufacturing enterprises. However, their uniform export poses a challenge due to the heterogeneity across multiple sources and the variety of export templates. Therefore, this paper proposes an intelligent extraction method for title block information (TBI) based on fuzzy matching. This method evaluates the geometrical composition of the title block, the logical relationship between information cells, and the diversification of the title block. It establishes a weighted intersection point matrix for the title block as the topological structure model. The fuzzy matching, based on the intersection matrix points, disregards the variations in the size and position of the title block and focuses only on the style that impacts the extraction of information. The matched title blocks are categorized, and the base table of each type determines the information distribution of the title block. A hash table is then created between the description cell and the value cell of information to accumulate the TBI. The method is integrated into the AutoCAD software and presented with an interactive program interface. Finally, an example is presented to further demonstrate the stability, reliability, and efficiency of the method and the system.