| 摘要: | 隨著國內臺灣地區近年來經濟發展進步，各項公共建設的設置不斷增加，這也意謂著土地使用資源改變快速。但相對地，卻沒有針對特殊應用的全面性土地使用資料來供各項建設整體規劃之需要。本研究即基於以上需求，探討如何產生臺灣地區全面性的土地使用資料。由於衛載的遙測影像涵蓋範圍廣大，為最佳的選擇，其中又以SPOT衛星影像的時間及空間解析力較能合乎需求。傳統上，若要從遙測資料獲得土地使用類型，大多採分類的方式。但由於衛星影像光譜解析度有限且由遙測影像所分類出來的是土地覆蓋的資料，若要得知實際的土地使用情形，不能單靠光譜資訊將其類別分出，要加入其他輔助資料才能求得。 本研究是以無線電網路規劃所需求之土地使用資料，根據其無線電傳播之特性，產生包括森林、水體、開放地、都市以及郊區五個類別的土地使用圖。研究顯示衛星影像的光譜分類在此應用會產生一些實際困難。例如，雲層及其產生的陰影會有遮蔽影響；森林類別會受到山區地形效應的陰影遮蔽影響；河床在河流枯水期會露出於水表面，使得河流類別會產生不連續的現象；衛星影像分類只能將建地類別分出，沒有辦法將都市以及郊區土地使用類別分出。所以本研究嘗試將SPOT衛星影像和地理資訊系統(geographic information systems，簡稱GIS)的向量資料相整合，以幫助監督式分類(supervised classification)訓練區的選取與土地類別資料作進一步的處理，克服其困難以便產生無線電網路規劃所需求的土地使用資料。 Coming with the progress in economic development in Taiwan recently, the number of various public constructions has continued to increase. This also implies that the resources for land use are changing fast. Relatively, there is, however, no comprehensive data of land use aimed at special application for the needs of integrated planning of all developments. Based on the above requirements, this research studies how to generate the comprehensive data of land use in the Taiwan area. Because the remote sensing image from the satellite covers a large area, in order to get the best results among all choices, the SPOT satellite image offers the temporal and spatial resolution that can best meet our requirements.Traditionally, if we want to get the various types of land use from remote sensing data, we usually adopt the method of classification. because the spectral resolution from satellite images are limited, and the classification generated from remote sensing images is data of land cover, if we want to know the actual way land is used, we cannot rely solely on the spectral information to make classification, additional auxiliary data are required. This research is using radio network to plan the required data of land use. According to the characteristics of radio broadcasting, charts of land use consisting of five classifications including forest, water, open space, urban, and suburban are generated. Previous studies have shown that the spectral classification of satellite images in this application will encounter some difficulties. For example, clouds and its shadows would have cover effects; the classfication of forest would be affected by the mountain geographic effects; the bottoms of rivers would emerge above water during drought seasons, making the classification of rivers (or water) inconsistent; satellite image classification can only classify the classfication of built-up land, and cannot classify the land use of urban and suburban. As a result, this research tries to combine satellite images from SPOT and the vector data from Geographic Information Systems (GIS) to faciliate the selection of training areas of supervised classification and further processing of land use classification, overcoming all difficulties so as to generate the charts of land use required by radio network planning. |
