基于灰度图像的压印凹凸字符质量检测研究  

Product Quality Inspecting On-line for the Pressed Protuberant Character

作　　者：曹建海[1,2] 路长厚[3] 李磊 

机构地区：[1]清华大学计算机科学与技术系 [2]深圳市现代计算机有限公司,广东深圳518057 [3]山东大学机械工程学院 [4]山东省济南信息市场

出　　处：《光电子．激光》2007年第1期89-92,96,共5页Journal of Optoelectronics·Laser

基　　金：山东省重点产业化资助项目(0203c06)

摘　　要：提出了直接在灰度图像上进行的压印字符质量检测的新方法。首先利用基于矢量和的最优圆心定位算法得到字符相似误差进行1次质量检测,然后以3点定位获得倾角误差进行2次质量检测。该方法是直接在灰度图像上提取检测特征,不仅保留图像的原始特征,克服了通常字符检测算法过分依赖于二值化算法及抗干扰性差等的弊病,而且快速,适合在线检测要求。实验表明,该方法不仅能检测不同灰度、位置和角度的压印字符,而且也适宜于有重叠、缺损等干扰的字符图像,正确检测率为97.27%。The pressed protuberant character is a refleetorized character on the difference of reflectance. The image quality of the protuberant character is lower than that of the general character on the chromatic difference of background and fore ground, So,the general character quality inspeetiing method based on the binary-scale is unfit for the protuberant character. To solve this problem,a new method for direct gray-scale feature extraction based on the ring projection algorithm and the vector sum for inspecting the quality of the pressed character is presented. The new method keeps integrity feature informa tion of the protuberant character and can overcome the main shortages of the traditional method on binary＇scale image, such as binarization algorithm, the lower performance of anti-jamming. The experimental results show that the proposed method adapts to detecting the variant gray scale, location and orientation of the character,and can yield an excellent performance under the conditions of noise and deformity.

关 键 词：灰度字符 质量检测 矢量和 特征抽取 

分 类 号：TP274.3[自动化与计算机技术—检测技术与自动化装置] TP391.43[自动化与计算机技术—控制科学与工程]