Automatic Dimensioning Of Engineering Drawings Based On Feature Recognition

The engineering drawing is a programmatic document for product design and manufacture.Dimensioning is the basis for manufacturing and inspection;drawing dimensioning is time-consuming and laborious,and problems such as missing labels,mislabels,or non-standard labels often occur;automatic dimensioning can reduce Designers burden,improve design efficiency.Therefore,it is of great practical significance to research and implement the automatic dimensioning method of drawing dimensions.Firstly,the semantics of the mechanical parts drawing is deeply analyzed.The engineering semantic information of different projection views is classified.The features and functions of engineering drawing are analyzed,and the constraint relations and structural features of the semantic information of the engineering drawing are defined,Feature recognition is necessary to automatically dimension engineering drawings.The feature model corresponding to each feature is constructed,the corresponding shape representation of each feature in the engineering drawing is analyzed,the relationship between the contour loops of each feature is elucidated,which lays the foundation for the identification of the subsequent feature loops.In addition,Describes the principle of the conditional random field model in machine learning and the learning method of conditional random field are described,namely the parameter training method of the model.Using the chained conditional random field in the conditional random field model,the part features in the drawing is identified.The identified loops are established based on the relationship between the feature contour loops,the summarized attributes of each loop.The characteristics of the relationship between a feature loop and the other loop,the establishment of attribute feature set and relationship feature set,the establishment of the feature of each feature,in order to estimate the model parameters.The feature labeling analysis was performed,and the labeling rules were established.Finally,the identified features were marked according to the rules.According to the above algorithm and research and using object-oriented programming techniques,the implementation and example tests are carried out on the JHCAD platform,which verify the feasibility and effectiveness of the presented method.