基于霍夫变换的铭牌OCR图像旋转矫正方法  

Nameplate OCR image rotate correction method based on hough transform

作　　者：张勇红[1] 

机构地区：[1]甘肃省电力公司,兰州730030

出　　处：《电测与仪表》2015年第8期125-128,共4页Electrical Measurement & Instrumentation

摘　　要：在对互感器铭牌图像进行扫描输入时,铭牌图像或多或少会出现一定程度的倾斜,这种图像的倾斜最终会导致其字符识别准确率下降。针对此问题提出一种基于霍夫变换获取图像倾斜角度,进而通过图像旋转矫正提高光学字符识别(Optical Character Recognition,OCR)准确率的方法:首先对原始图像进行二值化,进而获得铭牌的轮廓,再采用基于霍夫变换的方法获得铭牌中的水平线段,通过计算得到线段的水平倾斜角,利用此倾角对图像进行还原。实验结果表明,该方法能快速地计算图像的倾斜角度,提高了OCR识别准确率且准确率可达95%以上。In the process of scanning the transformer nameplate image, images may slant somehow, which may lead to declination of the accuracy rate of character recognition. A method to improve the accuracy of OCR based on Hough transform has been proposed. Firstly, we make the original image binaryzation, thus the outline of the nameplate is obtained, and then acquires the horizontal lines of the nameplate based on Hough Transform, and horizontal angle is obtained by calculation, we can restore the image through this angle. The experimental results show that it＇ s very fast to calculate slant angle of images and improve the accuracy rate of OCR above 95% when using this method.

关 键 词：OCR 字符识别 霍夫变换 旋转矫正 

分 类 号：TM93[电气工程—电力电子与电力传动]