A Highly-Efficient Constraint-Network-Based Approach for Symbol Recognition of Engineering Drawings

This paper converts the geometrical features of all the symbols in engineering drawings into one constraint-network, whose nodes contain the test conditions and edges denote the judgement order, and implements the new filtration recognition algorithm based on the hierarchical condition testing. This approach is guided by the thought of stepwise simplification, introduces multi-layer priorities, and optimizes the flow of network traversal through defining edge attributes cooperated with priority adjustment. By using the procedure of recording the indices of primitives that have successfully passed former condition test in certain nodes, the times of primitive traversal and judgement are reduced greatly. The approach of the paper is implemented in one architectural drawing recognition system, embodies high efficiency and powerful adaptability, especially for the situation of large quantity of symbol types or the case of complicated drawing.