頁籤選單縮合
題 名 | 布疋瑕疵檢測之影像處理理論探討(1)--共發矩陣特徵值最佳化=Theoretical Evaluation of Fabric Defect Images by Using Image analysis(Ⅰ)--Optimization of Characteristic Values in Co-occurrence Matrix |
---|---|
作 者 | 陳鴻仁; 劉致華; 連信仲; | 書刊名 | 紡織中心期刊 |
卷 期 | 10:3 2000.07[民89.07] |
頁 次 | 頁192-201 |
分類號 | 478.11 |
關鍵詞 | 共發矩陣特徵值; 布疋瑕疵影像; Characteristic value of co-occurrence matrix; Fabric defect images; |
語 文 | 中文(Chinese) |
中文摘要 | 本文係針對梭織胚布瑕疵分類之類神經網路的共發矩陣機率條件特徵值,做角度與距離的最佳化。主要是觀察正常 (無瑕疵)、斷經、斷緯、破洞、油汚、倂經、倂緯、稀弄與密路等瑕疵影像的共發矩陣特徵值:均勻性度量的二階矩、對比度、最大機率、、均勻度、相關係數、1階到5階元素差分矩、1階到5階元素逆差分矩,在角度0、45、90與135度時,特徵值與距雖位置之週期極值關係,作為計算共發矩陣機率條件特徵值的依據,采取最佳的特徵值,作為瑕疵分類之類神經網路系統的輸入參數,使瑕疵分類過程更快速,瑕疵分類結果更正確。 |
英文摘要 | In this paper, the characteristic values of co-occurrence matrix of fabric defect images to specify the optimum angles and distances. Were observed. The fabric defect images include normal area, broken end, broken pick, hole, oil stain, double end, double pick, filling bar and crack. The characteristic values of the images which include angular second moment, contrast, maximum probability, element difference moment, inverse element difference moment of order K, entropy, uniformity and correlation are then calculated . The optimum angles and distance of the characteristic values are evaluated and then input into a neural network for the classification of fabric defects. By using this method, the accuracy of fabric defect classification can be increased. |
本系統中英文摘要資訊取自各篇刊載內容。