Abstract
The class of drawings considered here is characterized by alternative instances of symbols and connections, includes flow charts and diagrams describing chemical plants, electrical circuits, and digital logic. After preprocessing of the scanned binary image, symbols and connection lines are separated using a few simple yet effective generic drawing properties, such as: (a) symbols contain mostly closed simple shapes, slant lines and open lines, and (b) connection lines contain mostly horizontal and vertical lines with each line end connected to a symbol. Accurate results are achieved in experiments on a set of 24 drawings drawn from four applications and scanned with resolutions varying from 75 to 300 dpi. Such high accuracy of segmentation can form the basis for symbol and connection recognition in any automated drawing understanding system.
Original language | English (US) |
---|---|
Pages (from-to) | 391-404 |
Number of pages | 14 |
Journal | Pattern Recognition |
Volume | 27 |
Issue number | 3 |
DOIs | |
State | Published - Mar 1994 |
Keywords
- Benchmark results
- Domain independence
- Engineering drawing analysis
- Flow drawings
- Loop analysis
- Segmentation
ASJC Scopus subject areas
- Software
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence