选票扫描图像识别中的快速定位方法

RAPID POSITIONING METHOD FOR SCANNED BALLOT IMAGE RECOGNITION

在基于OCR技术的选举投票系统中,如何实现图像识别中的快速定位仍然是一个重要问题。为了解决这一问题,在充分考虑选票设计并利用图像处理技术的基础上,提出一种选票扫描图像识别中的定位方法。该方法利用选票设计中记录的相关位置数据,将选票图像预处理后,找到一个确定的参考点,再提取选票设计中的位置数据,将它转化为相对于这个参考点的图像位置,利用该位置搜索表格线,从而确定要识别或要截图图像的准确位置。实验结果表明,这种方法的定位速度非常快,应用到选举投票中选票扫描图像的识别是可行的。

In OCR-based election voting system, how to achieve rapid positioning in image recognition is still an important issue. To solve this problem, a positioning method for scanned ballot image recognition is proposed based on taking full account the design of the ballot and utilising the image processing technology. The method uses relevant position data recorded during the ballot design, pre-processes the ballot image to find a definite reference point, then extracts the position data in the ballot design, converts it to an image position data relative to the reference point position, and searches the table lines with the position data, so as to determine the accurate position of the image to be recognised or captured. Experimental results show that this method has high speed in positioning, and is feasible to be applied to election votes for scanned ballot image recognition.