Study Of Key Techniques In The Printed Mongolian Characters Recognition

In the 1980s, research of Mongolian characters input methods was begun. Most of input methods were concentrated on the keyboard code. But research of Mongolian characters recognition was quite little. Under the circumstances, we proposed to research and realize a recognition system for multi-font printed Mongolian characters. The recognition system not only can provide a new input method, which is quick, highly efficient and intelligent, but also has a far-reaching meaning for inheriting and developing the minority culture as well as prompting the minority society further.The paper proposes a lot of improvements and innovations on the basis of the former research production. The main work is as follows:1. Automatic skew detection method of Mongolian document images. A skew detection method based on least square method for Mongolian document images is put forward in this paper. First, in a Mongolian document image, all the connected components of characters are searched. Then, characters are merged into character columns by relative positions between different characters. In the same character column, centroid of each character connected component is taken as an eigen-point and uses the least square method for straight-line fitting. The skew angle can be achieved from the slope of the straight line.2. Layout analysis method of Mongolian document images. A layout analysis method based on connected components for Mongolian document images is proposed in the paper. It is a combination of the bottom-up approach and the top-down approach. First, in a Mongolian document image, all the connected components of elements are searched. Then they are classified by their size (length or area), and we can get different elements (characters, images, tables, etc.) of document image. And finally, we put the connected components of characters into character columns and blocks.3. Research and realization of Mongolian letter segmentation method. A letter segmentation method based on backbone is presented in this paper. This method supplies pre-condition for feature extracting and feature matching.4. Feature selection of the printed Mongolian characters. Many features of Mongolian characters are selected, which include coarse classification features and fine classification features.Experiment proves that they can fulfill the task of the printed Mongolian characters recognition. The performance and the automation of the system have achieved highly levels.