电脑知识与技术
電腦知識與技術
전뇌지식여기술
COMPUTER KNOWLEDGE AND TECHNOLOGY
2014年
23期
5496-5500
,共5页
数字识别%最近邻算法%外接矩形%置信度
數字識彆%最近鄰算法%外接矩形%置信度
수자식별%최근린산법%외접구형%치신도
numerical recognition%nearest-neighbor algorithm%External rectangle%confidence level
介绍了一种简易的LED数字识别的算法。与传统的数字识别方法相比较,识别精度较高。该算法对采集到的4个数字字符二值化后的图像用最近邻算法,并对最近邻得到的结果的置信度进行评估实现数字的识别。利用该方法时,通过对置信度的评估,只有4个数字都具有较高置信度的识别结果才能被接受,否则单独对某个字符重新细化处理,大大提高了识别的精度。减少因镜头模糊,倾斜,光照等因素而造成采集得到的图像的中字符发生的旋转,粘连,部分缺失而引起的误识。在以此算法为基础对256幅图片,每幅图片上4个字符的数字同时进行识别测试,正确识别率达到97.66%,误识率仅为2.34%。实验表明,该算法在识别精度、抗干扰性方面表现良好。
介紹瞭一種簡易的LED數字識彆的算法。與傳統的數字識彆方法相比較,識彆精度較高。該算法對採集到的4箇數字字符二值化後的圖像用最近鄰算法,併對最近鄰得到的結果的置信度進行評估實現數字的識彆。利用該方法時,通過對置信度的評估,隻有4箇數字都具有較高置信度的識彆結果纔能被接受,否則單獨對某箇字符重新細化處理,大大提高瞭識彆的精度。減少因鏡頭模糊,傾斜,光照等因素而造成採集得到的圖像的中字符髮生的鏇轉,粘連,部分缺失而引起的誤識。在以此算法為基礎對256幅圖片,每幅圖片上4箇字符的數字同時進行識彆測試,正確識彆率達到97.66%,誤識率僅為2.34%。實驗錶明,該算法在識彆精度、抗榦擾性方麵錶現良好。
개소료일충간역적LED수자식별적산법。여전통적수자식별방법상비교,식별정도교고。해산법대채집도적4개수자자부이치화후적도상용최근린산법,병대최근린득도적결과적치신도진행평고실현수자적식별。이용해방법시,통과대치신도적평고,지유4개수자도구유교고치신도적식별결과재능피접수,부칙단독대모개자부중신세화처리,대대제고료식별적정도。감소인경두모호,경사,광조등인소이조성채집득도적도상적중자부발생적선전,점련,부분결실이인기적오식。재이차산법위기출대256폭도편,매폭도편상4개자부적수자동시진행식별측시,정학식별솔체도97.66%,오식솔부위2.34%。실험표명,해산법재식별정도、항간우성방면표현량호。
This paper introduces a simple algorithm of LED digital recognition. Compared with the traditional method of digital recognition, this recognition accuracy of this method improved. Using the nearest neighbor algorithm, we recognize 4 digital character of the collected by camera at the same time. And only the high confidence of the nearest neighbor results can be accept-ed. Otherwise we can process the character which is low confidence individually again. In that way, we can greatly improves the recognition accuracy. Using this method, we can reduce the inclination for lens blur, illumination and some other factors, result-ing in the occurrence of the character image rotation, adhesion, partial loss caused by misunderstanding and rejection. By testing 256 images with 4 numeric characters every image ,the finally correct recognition rate is up to 97.66%, the error rate is only 2.34%. Experiments show that the algorithm performs well in terms of accuracy, anti-interference.