Study on line recognition in scanned engineering drawings

In order to enlarge the processing level with macrocosm idea, a classified recognition method is proposed, it classifies a bar-like block structure, analyses its topological adjacent relationship and recognizes lines by imitating uman vision process. In the block structure, a rational set in which all pixels have a common character retrieve the colony characteristic of related pixels, individual pixel has not such geometrical features, so the structure is a higher representation of individual pixel. After classification the blocks are organized into chains, lines (including arcs and curves) are extracted. In cross region of lines, the algorithm modifies only Bar-like blocks instead of dealing with points which are near intersection area. The results demonstrate the effectiveness.