摘要： 针对印刷体维吾尔文文字识别系统中的文字，尤其是连体段字母的切分这一难点问题，提出一种改进的投影切分方法。采用将基线区域像素点置白(像素点索引值置为255)，并对其进行垂直投影，取投影值为0的间隙中间位置作为候选切线位置，根据行高度与空白间隙宽度比的统计数据均值设置阈值，以此消除误切分。实验结果证明，该方法可有效提高连体段切分的正确率。

关键词: 印刷体维吾尔文, 光学字符识别系统, 连体段, 基线, 积分投影, 切分

Abstract: For the text in the print Uighur character recognition system, especially for the difficult issue of conjoined section letter segmentation, an improved projection segmentation method is put forward. The index value of the baseline region pixels set to white(set the pixel index value to 255) is adopted, and its vertical projection is made. The projector to a value of 0’s gap is taken as a candidate tangent position. The threshold value is combined with the line height and width ratio of the blank gap statistics data’ average value is set to eliminate false segmentation. Experimental results prove that the method makes the piece paragraph segmentation of the correct rate further improved.

Key words: print Uyghur, Optical Character Recognition(OCR) system, conjoined section, baseline, integral projection, segmentation

中图分类号: 

• TP391.41