摘要
针对基于水表图像的自动抄表,提出满足实时性要求的水表字符自动识别算法。首先,将真彩色图像灰度化和标准化处理;然后,通过字符分割得到独立的单个字符;接着,采用自适应阈值的二值化方法将灰度图像转换为二值图像,再对图像进行去噪和两次位置调整操作,将字符图像裁剪为模板大小并将字符提取到图像中央;最后,对图像进行细化、特征提取和字符识别。基于三类图像样本的实验结果表明,该算法整体的字符分割率可达97.33%,整体识别率达到94.00%,平均每幅图像的处理时间为0.15 s,能满足实时性要求。
A water meter character automatic recognition algorithm that can meet the real-time requirement is proposed for automatic meter reading based on the water meter image.The true color image is grayed and normalized,and then the indepen-dent single character is obtained by means of character segmentation.The binarization method based on the adaptive threshold is adopted to convert the gray image into the binary image.Denoising and two times of position adjustment operations are performed for images.Character images are cut into the template size,and characters are extracted into the center of images.Image detail-ing,feature extraction,and character recognition are conducted.The results of the experiment based on three types of image samples show that the algorithm has an overall character segmentation rate of 97.33%,an overall recognition rate of 94.00%,and an average processing time of 0.15 s for each image,which can meet the real-time requirement.
作者
陈英
李磊
汪文源
徐嘉伯
杨丰玉
CHEN Ying;LI Lei;WANG Wenyuan;XU Jiabo;YANG Fengyu(School of Software,Nanchang Hangkong University,Nanchang 330063,China)
出处
《现代电子技术》
北大核心
2018年第18期99-103,共5页
Modern Electronics Technique
基金
江西省自然科学基金(20161BAB212034)
江西省教育厅项目(GJJ160692)
国家自然科学基金(61762067)
国家自然科学基金(61501217)
国家自然科学基金(61662049)~~
关键词
水表字符识别
字符分割
自适应二值化
位置调整
自动抄表
自动识别
water meter character recognition
character segmentation
adaptive binarization
position adjustment
automatic meter reading
automatic recognition
