本文選題：光纖預警　切入點：模式識別　出處：《天津大學學報(自然科學與工程技術版)》2014年12期 　論文類型：期刊論文

【摘要】：由于傳統(tǒng)模式識別方法存在過學習、訓練時間長等缺陷,不能滿足光纖預警系統(tǒng)實時在線監(jiān)測的要求.相關向量機能夠克服傳統(tǒng)方法的缺點,識別精度高,向量機個數(shù)需求少,因此,將相關向量機應用于光纖預警系統(tǒng)模式識別中,采用小波能譜和小波信息熵的特征提取方法,在測試階段采用有向無環(huán)圖的方法進行多類識別.通過對威脅管道安全的事件進行實驗,識別精度達到92.67%,向量機個數(shù)只有2個,驗證了相關向量機方法應用于光纖預警系統(tǒng)的可行性和有效性.
[Abstract]:Because of the shortcomings of traditional pattern recognition methods, such as learning and long training time, it can not meet the requirements of real-time on-line monitoring of optical fiber early warning system. Correlation vector machines can overcome the shortcomings of traditional methods, and the recognition accuracy is high, and the number of vector machines needs less. Therefore, correlation vector machine is applied to pattern recognition of optical fiber early warning system. Wavelet spectrum and wavelet information entropy are used to extract features. In the testing stage, the method of directed acyclic graph is used to identify many kinds of information. Through the experiments on the events threatening the safety of pipelines, the recognition accuracy is 92.67, and the number of vector machines is only 2. The feasibility and effectiveness of the application of correlation vector machine in optical fiber early warning system are verified.
【作者單位】： 天津大學精密測試技術與儀器國家重點實驗室;
【基金】：國家自然科學基金資助項目(61240038)
【分類號】：TP212;TN911.7

【參考文獻】

中國期刊全文數(shù)據(jù)庫 前1條

1 袁浩東;陳宏;侯亞丁;;基于全矢小波包能量熵的滾動軸承智能診斷[J];機械設計與制造;2012年02期

【共引文獻】

中國期刊全文數(shù)據(jù)庫 前6條

1 李浩;董辛e，

本文編號：1618006