Research On Automatic Recognition Technology Of Electrical Wiring Drawings

The electrical wiring drawing is a standardized representation of the equipment layout and connection of power grid plants and stations.Power system technicians need to redraw and update a large number of image files of electrical wiring diagrams.The use of automatic image recognition technology to convert unstructured drawing images into structured data can help improve the work efficiency of technicians and increase the value of existing electrical wiring diagrams.Therefore,it is necessary to study the intelligent recognition method of image format electrical wiring diagrams.Electrical wiring drawing is a special kind of complex document image.Based on the summary of the research status of automatic recognition of various types of complex document images,this paper presents an electrical wiring drawing automatic recognition method EDRS(Electrical Drawings Recognition System)based on deep learning,To realize the detection and recognition of the electrical graphic element and symbol-text label in the electrical wiring diagram and the extraction of the relationship between different electrical elements.The main research work of this paper is as follows:(1)Aiming at the problem that multi-scale electrical primitive symbols and text annotations in large-resolution image files are difficult to accurately locate under the influence of complex interference factors such as blur and distortion,this paper constructs an electrical primitive symbol detection model based on Faster R-CNN.The introduction of the feature pyramid makes full use of the multi-level features of the image,accurately recognizes multi-scale electrical primitive symbols,and solves the problem that electrical primitive symbols are difficult to accurately identify under the influence of complex interference,and small-scale images are difficult to accurately locate.This paper also constructs a CTPN-based electrical text annotation detection model and a CRNN-based electrical text annotation recognition model.The model is improved by considering the differences and sequence characteristics of electrical text annotations of different lengths to enhance the recognition effect of electrical text annotations.(2)Aiming at the problem that various electrical elements are densely distributed in complex network diagrams,and the connection relationship is difficult to accurately identify,this paper constructs a convolutional neural network-based electrical graphic element symbol wiring endpoint recognition model to realize the connection of connecting lines and graphic elements Point location,design the electrical connection line recognition method based on image processing technology,combine the position information of electrical primitives,electrical connection lines and electrical text annotations to realize the extraction of the positional relationship between various electrical elements.(3)In view of the small difference in the shape of the group of primitives composed of independent primitives in the electrical wiring diagram,it is difficult to effectively extract the structural features and accurately identify the more difficult problems.The vector sequence representation method of electrical graphic element groups,constructing a one-dimensional convolutional neural network to realize the type recognition of electrical graphic element groups,solves the problem of difficult structure description and difficult recognition of graphic element groups.This paper uses image files of real electrical wiring diagrams for testing.The experimental results show that the method designed in this paper can effectively identify various electrical elements in electrical wiring drawing and their positional relationships,and is robust to complex interference factors.