复杂背景下航检视频字符的识别算法与应用  

Recognizers and Applications of Characters for Aeronautical Detecting Videos Under Complicated Background

作　　者：陈晓东[1] 杨伟旗 关鑫 汪洋 

机构地区：[1]内蒙古工业大学轻工与纺织学院,呼和浩特010080 [2]内蒙古超高压供电局,呼和浩特010080

出　　处：《内蒙古工业大学学报（自然科学版）》2017年第1期64-68,共5页Journal of Inner Mongolia University of Technology：Natural Science Edition

摘　　要：传统的OCR技术在复杂背景视频中的字符识别方面,其识别结果的准确率不尽如人意。为此,本文通过改进及优化传统识别算法,研究出一种新的复杂背景下的字符识别算法。该算法借鉴统计模式识别的思路,通过大量实验得出每个字符的特征规律,并形成公式,实现了字符的识别。然后以输电线路航拍图像为例,设计开发了字符识别系统。通过系统运行表明,采用本文算法识别字符的准确率为99.4%,优于开源Tesseract-OCR接口函数方法及样本训练方法,对于复杂背景中的字符识别具有较高的准确性和广阔的应用前景。The traditional OCR technology＇s accuracy rate of recognized results are disappointed in character recognition videos of complicated background. Therefore, we developed a character recognition algorithm applicable for complex background through improving and optimizing traditional recognition algorithm. The algorithm used statistical pattern recognition to get different weight coefficient of the characters and characteristics of statistical matrix by a large number of experimental training. Then we built a subspace characteristic curve with a characteristic area as the abscissa and the assessed value as the ordinate. Then we did a similarity match between ample curve under test and curve in the library similarity matching repository to achieve character recognition. Using this algorithm for processing transmission line aerial image of character recognition system indicates that the accuracy rate of recognizing characters can be 99.4; by our algorithm. This algorithm is better than open-source TesseractOCR interface function and sample training, and ithas a good accuracy and promising application for character recognition in complex background.

关 键 词：OCR 复杂背景 字符识别 统计模式识别 航检视频 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]