Extraction and Recognition of Device Graphics in Process Flow Diagram
Wei Shao, Guangze Wang, Hongliang Xi, Lei Feng · 2020 IEEE 9th Data Driven Control and Learning Systems Conference (DDCLS) · 2020
Design analysis on process flow diagram can find the logical error and improve efficiency. Existing software used to drawing the process flow diagrams (such as AutoCAD or Visio) does not have recognition and analysis function. For the images output by software, this paper describes the method of extracting and recognition the device graphics representing the valves, gauges, etc. Since the sizes of same device graphics may be different, the device graphics have to be extracted and normalized. In extraction, the points which are obtained by convolving the image with gradient operators, are used to positioning the device graphics. Considering the difficulty of extraction and recognition, the devices graphics are classified by shape. The device graphics with a circle are extracted by circle detection and normalized according to radius. The other device graphics are extracted, and normalized by the bounding rectangle after contours shrink. In recognition, the improved template matching and the recognition method based on structure are used separately to recognize different device graphics. After experimentation, the method described in this paper can significantly improve the recognition rate of various types of device graphics.