Recognition and Extraction of Information from Image-based Tables for Electric Power System Operation and Maintenance

During the operation of power system, massive documents are generated, which usually contain complex tables. The review and examination of these documents by manual are very time consuming. To improve efficiency, this paper proposes a recognition and information extraction method from complex tables in document images. First, features of typical tables are analyzed and the concept of minimum recognition unit is proposed. Then, a skew image correction method is introduced. Afterwards, the opening and closing operations in mathematical morphology are applied to detect the border of the tables and segment the image into minimum recognition units. Finally, text recognition technique is applied to specific minimum recognition units. The performance of the proposed method is tested with different types of images under different conditions. The results verified the effectiveness of the proposed method.