頁籤選單縮合
題 名 | 濁水溪河口揚塵潛在區位地覆類別判釋之研究=A Study of Landcover Pattern Recognition at the Dust Emission Potential Areas in the Estuary of Jhoushei River |
---|---|
作 者 | 林昭遠; 李承寯; 莊智瑋; | 書刊名 | 水土保持學報 |
卷 期 | 41:2 2009.06[民98.06] |
頁 次 | 頁125-138 |
分類號 | 434.273 |
關鍵詞 | 倒傳遞類神經網路; 植生指標; 紋理因子; 影像判釋; Back propagation neural network; Vegetation index; Texture factors; Image classification; |
語 文 | 中文(Chinese) |
中文摘要 | 河床裸露地易受東北季風吹襲而產生揚塵,危害鄰近區域空氣品質,如何有效快速劃定河 口揚塵潛在發生區位及加強復育管理極為重要。本研究使用衛星影像輔以植生指標與紋理因 子,萃取影像中之波段資訊並運用於植被、裸露地及水域之判釋。利用原始波段加入NDVI 及 CMFI 等兩種植生指標,並添加紋理因子,最後,將其優選後之植生指標結合原始波段及紋理 因子帶入倒傳遞類神經網路分類器進行地覆類別之判釋。結果顯示,以原始波段合併植生指標 CMFI 有較佳的判釋能力,且藉由加入之紋理因子,更可再提昇影像分類成果之精度。經探討 後發現,加入之紋理因子以3×3 之移動視窗大小之判釋精確度最高,各期判釋成果之Kappa 係 數均達0.66 以上。 |
英文摘要 | Due to the characteristics of flat terrain, huge bare soil in drought season, and the monsoon effect, the estuaries of rivers in central Taiwan are susceptible to serious wind erosion. Large amounts of dust emission to inland decrease the living quality and affect the human health nearby in the monsoon season. Therefore, it is important to delineate the potential areas where are mostly susceptible to wind erosion on the riverbed and to make some vegetation strategies for the areas of dust emission. The supervised back propagation neural network technology was employed to compare the accuracy of image processing by using the vegetation index and/or the texture factors which derived from SPOT satellite imagery for the bare land classification. Results show that the image processing with original bands plus vegetation index CMFI has the better interpretation and the accuracy of image classification can be further improved by adding the calculation of texture factors. The best moving window size is 3×3 for the texture factors calculation. |
本系統中英文摘要資訊取自各篇刊載內容。