頁籤選單縮合
題名 | 利用紋理因子改善影像分類準確度之研究=A Study of the Accuracy Improvement for Image Classification Using Texture Factors |
---|---|
作者 | 莊智瑋; 林昭遠; 陳正湘; | 書刊名 | 水土保持學報 |
卷期 | 41:2 2009.06[民98.06] |
頁次 | 頁153-167 |
分類號 | 440.98 |
關鍵詞 | 紋理因子; 影像分類; 影像灰階值; 灰階共生矩陣; 移動視窗; 植生指標; GLCM; Texture factors; Image classification; Image gray level; Moving window; Vegetation index; |
語文 | 中文(Chinese) |
中文摘要 | 近年來科技日益進步,多光譜及高空間解析影像取得容易,對於環境變遷、災害監測及土 地利用分類等應用已普遍備受重視,傳統影像分類乃利用影像灰階值之差異進行歸類,但影像 上除影像灰階值之差異外更有紋理上之不同,一般紋理計算以灰階共生矩陣最為常見。因此, 本研究以九份二山為研究樣區,除採用傳統之影像灰階值差異進行分類外,更加入6 種紋理因 子及不同移動視窗進行探討,探究不同組合方式對影像分類精確度之差異。結果顯示,以原始 波段加入植生指標及3X3 移動視窗所得分類結果最佳;在移動視窗大小部分,發現原始波段加 入移動視窗愈大之紋理影像,精確度卻逐漸降低,可能係移動視窗愈大雜訊愈多所致。研究所 得結果可供未來進行影像分類之參考。 |
英文摘要 | Science and technology is rapidly progressive in recent years. Multiple spectrum and high resolution image are commonly used in the study of environmental change, hazard monitor and land use classification. The spatial distribution of an image’s spectrum is not only different in gray value but also in texture. Gray Level Co-occurrence Matrix can be used for texture calculation, therefore; this study is focused on using the traditional image classification couples with six texture factors and different moving window size to discuss the effect of accuracy improvement in the study area of Chiufenershan. The result shows that the combination of original bands, vegetation index and texture factors with 3X3 moving window size having the best performance in land cover classification. For the moving window treatment, noise will significantly decrease the accuracy of classification due to increasing the window size. The results of this study can be as the reference for land cover classification. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。