An Automatic Recognition System of Assembly Drawings

tion rate. It is not an easy task to make and modify software Automatic recognition of mechanical assembly drawings is and symbol dictionaries, according to the types of the a difficult problem in drawing recognition. In this paper, a diagrams. heuristic recognition method 'serial numberleader line-part A more recent interest in drawing recognition is that of symboln is described to recognize assembly drawings, A trackw~ing to design systems which embody knowledge about the ing algorithm is proposed to a variety of leadcr lines, basic structures of defferent kinds of graphic symbols and use find the corresponding part symbols. A tracking-matching this knowledge to analyze and identify the different compoalgorithm is built to track and assembly drawings in nents of a drawing. Such a system would tie together various terms of a generating-testing process of primitives, Because alimage analysis, like thinning, edge gebra expression of the part symbol lots of semantic fcaWmcntation, filtering, etc. along with s high-level control tures about the part symbol, the tracking process is simultaStructure that interprets the drawing image with the help of ncous with the process, The experimental results these image processing operations. A knowledge-bascd drawshow that our method is effective and isn't affected by comp~cx h g recognition system involves the use of several types of degree of assembly drawings. knowledge including visual, spatial and linguistic. A knowledge-based system that directs the classification of the different entities on a drawing image and decides when an I . INTRODUCTION unambiguous classification of all the relevant entities has been Drawing recognition, which converts hand-drawn data into computer accessible codes, is one of the important branches in pattern rccognition. Automatic assembly factories use 'assembly drawings", which inform the species of parts and the interconnections among the parts, to accomplish mechanical assembly tasks. Because there are complicated connection relations of parts in mechanical assembly drawings, recognition of assembly drawings is very difficult. So far, some methods have been developed in order to achieved, is one of major goals in the field of drawing rccognition. Here, an automatic rccognition system ARSAD of assembly drawings is introduced, which can effectively utilize mechanical, graphic semantic knowledge to recognize assembly drawings. A mechanical assembly drawing consists of six kinds of elements: serial numbers, leader lines, a list of parts, graphic part symbols, a outline, and character strings. 1) serial numbers: they indicate Nos. of part symbols, and are used to indicate which graphic symbols leader lines point recognize equipment location maps, network diagrams, electronic circuit diagrams, chemical plant engineering 2) leader lines: they indicate the relationships betwcen seridrawings[,-41. [I], ~~~~i proposed a new template matchal numbers and graphic symbols, or configuration of a outline. ing method which is very fast for binary images, this They are represented as two connected stragiht lines. mcthod, the image is shifted by specified distances instead of 3) list of parts: it indicates the relationships between serial moving the template. Among the shifted images,logical operanumbers and Part names, quantity of parts, specification of tions arc performed on the entire image area. Thcse operations can be performed very quickly with the raster operation in a workstation. In [2], Macda developed the method of symbol recognition, according to the sequence of the following steps: a) key constituent detection; b) constituent detection; c) symbol determination; d) reclassification of remaining lines. In [3], Fahn presents a topology-based component extractor for the understanding of electronic circuit diagrams. Thcse earlier systems don't utilize the semantic knowledge, what is expressed in the diagram. And this limits the possibility to improve rccogniparts, etc. 4) graphic part symbols: they indicate certain spccific part symbols which stand for general mechanical parts, and arc rcpresented as fixed graphic symbols. In our ARSAD system, we can recognize all kinds of assembly drawings composed by 31 part symbols shown in Fig.1. 5) outline: it indicates the external form of the assembly drawing. There exists an unique outline in every assembly drawing. 6) character strings: they indicate comments in list of