Intelligent Extraction Technology Of Key Information In Engineering Drawings

With the wide application of cartographic software and plenty of accumulated engineering drawings,extracting and retrieving information from engineering drawings quickly have become an urgent problem to be solved.At present,the main method to search drawings in a database is the manual comparison,which is cumbersome and inefficient.Therefore,it will improve the management efficiency of engineering drawings greatly via the rapid key information extraction in an engineering drawings and the establishment of drawings information database and similar drawings search system.Based on this background,this thesis studies the technology of using artificial intelligence technology to extract key information from engineering drawings quickly and automatically.The main work is to extract the title bar information and marking dimension information in engineering drawings,and designs a set of information extraction and query system.The main research work of this thesis is as follows:Firstly,in this thesis,the method to extract the title bar information of engineering drawings is studied,which involves both text region location and text recognition two aspects.In view of the fixed format of the table,the key information is positioning analyzed by the form of splitting table cells.The experiment shows comparing with the traditional recognition rate,a character recognition model based on convolutional neural network,helps to improve the recognition rate of character effectively.In addition,it achieves the automatic extraction of title bar information.Secondly,considering the inefficiency of traditional feature-based marking dimension detection method and the fact that the extraction of annotation information is the result of the combination of characters and constituent element.in the thesis,it is novel to use deep learning method to detect both the marking dimension and the character at the same time.Through experiments,it is found that this method can locate the dimension information accurately.The recognition and integration of labeled digits after positioning achieves the effective extraction of the dimension information.Finally,aiming at the need of similar drawings search,the whole process of automatic extraction of key information of drawings is designed to extract the title bar information and label information of drawings quickly and intelligently,and import them into the database.On the basis of the accomplished database,a drawing search system is designed.The system extracts the key information from drawings automatically,and then the most matched drawings will be outputted after comparing with the database.This system eliminates the cumbersome process of manual search and improves the efficiency of the search greatly.