Automatic recognition of reinforced-concrete CAD drawings is of great significance to the digital transformation of the construction industry, as it can dramatically improve the efficiency of BIM modelling, structural analysis and quantity take-off. This paper studies the automatic recognition technology for RC CAD drawings. For the reinforcement drawings of walls and columns, a series of algorithms are proposed, including the automatic identification and separation of columns and walls, the identification of various types of reinforcing bars corresponding to columns from the reinforcement table, and the alignment of the reinforcing bars to the column body. It is verified by examples that the algorithm in this paper can efficiently extract the geometric contours of various columns (including inclined and special-shaped) and walls and their corresponding column reinforcements and accurately align them to the columns. A complete structured data is formed, which can be connected to all kinds of digital applications, and has significant engineering application value and broad industrialization prospects.