面向药瓶标签的字符定位方法研究

针对药瓶标签的特点,提出了一种数学形态学和投影特征相结合的药瓶标签字符定位方法。该方法首先对药瓶标签图像进行预处理,减小圆柱面造成的字符畸变程度,对处理得到的图像进行二值化和边缘检测等。其次对图像进行多次数学形态学处理,使图像分割成几个连通区域,根据先验知识以及轮廓跟踪方法得到字符所在的大致区域。最后利用水平投影和图像分割方法将粗定位得到的图像分割成三个单行字符,完成对字符区域的精确定位,便于后续的单个字符分割和识别。实验结果表明,该方法能对药瓶标签进行有效的定位。

Character Positioning Methods of Medicine Bottle Label

According to the characteristics of medicine bottle labels, a method of character positioning which combines mathematical morphology with projective characteristics is provided. According to the method, firstly, the images of the medicine bottle labels are preprocessed so as to reduce the distortion extent of characters caused by cylindrical surfaces and the processed images will be performed binary and edge detection etc. Secondly, the images are performed mathematical morphology process many times to be divided into several connecting regions. According to the prior knowledge and contour tracking scheme, the characters area can be got. Finally, the images can be segmented into three separate characters by the level projection and image segmentation method. The accurate positioning of target characters is completed so as to segment and recognize succeeding single character. The experiment results show that the method is effective on the medicine bottle label position.