台灣地處環太平洋地震帶，是典型的板塊碰撞下產生之大陸邊緣島嶼，造山運動迄今仍在持續的進行，引發台灣密集的地震活動；台灣西隔台灣海峽連接歐亞板塊，東臨太平洋，位處於世界季風最發達區域內，盛行風隨季轉變而有顯著改變，梅雨季與颱風帶來旺盛雨量，亟易造成山區土石流等地質災害。 衛星雷達由於俱全天候特性，可及時監測提供災害資訊已經成為環境監測的重要工具。由於重複軌道拍攝幾何關係相似，像元攝像入射角度接近，在地表起伏地區，雷達影像幾何變形特徵一致，常見雷達影像失真現象如疊置、陰影等。本研究利用不同時期、有相同地區重疊的雷達影像，經過以多尺度匹配法(Multi-scale matching)為主的處理過程，以及濾除雜訊的步驟，有效地在雷達影像上找到均勻分布且精確可靠的共軛點，進行影像套疊，提供後續的山區崩塌偵測研究。 建置衛星雷達影像自動化套疊與變遷分析，可減少人為誤差及提昇效率。研究個案為台中縣境內山區的2006年612水災前後期合成孔徑雷達影像，即時偵測大甲溪上游的知馬漢山地區。隨著公尺級超高解析衛星雷達Radarsat-2與TerraSAR-X即將升空運作，本研究成果將可滿足全天候山區崩塌監測實務需求。 Taiwan is the Continental island of collision between continent and to lie in the Circum-Pacific seismic zone so that epeirogenic movement is still proceeding violently, account for earthquakes hourly. In addition, Taiwan is a partition between the Taiwan Strait to connect with Eurasia and the Pacific Ocean, it is in the growthful rains zone. Typhoon and plum rain account for torrential rain and cause easily geologic disaster as landslide. Satellite Radar has the all-weather property and repeat observation, it can supply geographical coordinates of disaster by monitor immediately. Because repeat-pass observation, the geometry relationship is similar. Besides, the geometric characteristics of Synthetic Aperture Radar image, like layover and shadow, are also similar due to the similar look angle of camera. In this study, we used the SAR images have multi-temporal and overlap to handle with the multi-scale matching and dispeckle. Finally, we found efficiently the tie points scattered uniformly in the SAR image and use this to registrants images. In this study, the multi-scale automatic registration of satellite radar images can reduce artificial error and increase efficiency. Such case is located at Taichung mountainous area in 2006, we can detect change in time. Recently, the high resolution satellite SAR like Radarsat-2、TerraSAR-X will hurl and work, automatic registration method can monitor land cover change sensitively.
