Methods of collecting rebar quantities information in bridge design drawings based on deep learning
There exists rebar quantity information in bridge design drawings,which is used in the construction process for rebar alloca-tion work,and at the same time,this information can provide design experience in the design of new projects,however,this rebar quantity information is usually saved in paper drawings in the form of tables,which is inconvenient for access and reference.Based on this,a method is proposed to summarize the rebar quantity from bridge design drawings.Firstly,in the backbone network of YOLOv7,the convolutional attention module(CBAM)attention mechanism is incorporated to obtain more detailed features,and the multi-branch convolutional RFB module is introduced in the head structure to improve the feature expression ability of the model for small targets,object detection of normal tables and title block of bridge design drawings based on improved YOLOv7;secondly,the table recognition function provided by PP-StructureV2 is used to identify the norma tables and title block in the drawings;after summarization and vali-dation,an Excel table containing all the rebar quantity information is finally generated to realize the summarization of rebar quantity in-formation in bridge design drawings.Experimental studies were conducted using the bridge design drawings provided by Anhui Transport Consuling&Design Institute Company Limited.The experimental results show that the F1 Score of the improved YOLOv7 algorithm provided in this paper can reach 98.35％,which is 0.86％higher compared with the original YOLOv7 algorithm,and it can meet the requirements of detecting the common table as well as the title block table from the bridge design drawings.
NETL
NSTL
万方数据
