摘要
针对监控屏幕中的数字字符提出了一种高效的识别算法。该算法利用字符图像的欧拉数、凹陷区、水平和垂直穿线等组合特征完成级联分类,无须对待识别字符进行规整、细化和轮廓提取处理,降低了算法复杂度,减少了因细化变形、轮廓断裂引起的误识和拒识。在以此算法为基础实现的监控信息自动采集与记录系统中,对5 000多个屏幕显示数字字符进行识别测试,平均每秒处理125个数字,正确识别率达到98.70%,误识率仅为1.30%。实验表明该算法在处理速度、识别精度、抗干扰性方面表现良好。
An effective recognition algorithm is proposed for digits on monitor screen. By using the features of euler number, concave field, horizontal and vertical crossing Line, digits can be classified and recognized. The algorithm doesn＇t needs any regularization, thinning and outlining operations on the digits image. So it achieves low memory and computation. With the proposed method a monitor information catching and record system is realized. More than 5000 digits are involved in the expe- riments and a eonelusion is drawn that the average reeognition rate is 98.70%, the error rate is 1.30%, and the reeognition speed is 125 digits per seeond. Experiments show that the algorithm is fast with high reeognition rate and strong anti-interference ability.
出处
《计算机应用研究》
CSCD
北大核心
2006年第10期172-173,182,共3页
Application Research of Computers
基金
陕西省自然科学基金资助项目(2004F32)
关键词
数字识别
欧拉数
凹陷区
穿线
Digits Recognition
Euler Number
Concave Field
Crossing Line
