攝影測量(Photogrammetry)是現今應用範圍十分廣泛的技術。影像 的測量應用於災害防治的監測已行之有年，與傳統的監測技術相比， 攝影測量具有高機動性、成本低且安全性佳的優勢。隨著科技進步， 除了光學影像分析外，熱影像的監測也開始應用於邊坡災害的防治。 熱影像與可見光影像不同，並不能得到詳細的物體位移資訊，但是可 以提供物體的溫度資訊供監測者進行區域性的追蹤。近年來研究指出， 熱影像可以透過溫差分析隱藏於邊坡表面後潛在的不穩定區域。因此 若能透過光學影像的分析保留邊坡的幾何資訊，又輔以熱影像資訊追 蹤邊坡中隱藏的不穩定區域，可以提供更具整體性的監測數據。 本研究的主要目的是利用電腦視覺技術融合光學及熱影像，將熱 影像的溫度資訊套疊至光學影像上進行邊坡資訊的分析，並研發現地 設備將該方法應用於實務上。除了可以維持原本邊坡位移的分析外， 亦能透過熱影像進行溫度差異分析，分析邊坡後方是否有潛在的危險 區域以利未來規劃與追蹤。本研究在開發完成現地儀器設備後，會以 逆向坡的縮尺模型試驗進行監測儀器的評估並以中央大學後方邊坡 為例，透過現地架設監測系統，從多期的影像進行邊坡位移的分析， 並透過後期處理的三維資訊模型實現本研究整體系統的監測流程。根 據實驗成果，精度可以達到 3 公分以內，尚可精進。其後續成果可以 提供邊坡災害預警、防治與災害後規劃，提供更有效的決策數據。;Photogrammetry is a technology with a wide range of applications nowadays. Compared with traditional technology, photogrammetry has the advantages of high mobility, low cost, and safety. With the advancement of technology, thermal image monitoring has also started to be applied to the prevention of slope disasters. Unlike optical images, thermal images do not provide detailed geometric information about the object, but they can provide temperature information instead for regional tracking by the monitor. Recent studies have shown that thermal images can be used to analyze potentially unstable areas hidden behind the slope surface by temperature difference. Therefore, if the geometric information of the slope can be retained through optical image analysis and supplemented with temperature information to track the hidden unstable areas in the slope, it can provide more comprehensive monitoring data. The main objective of this study is to use computer vision technology to fuse optical and thermal images for analysis and develop a field instrument based on the method. In addition to maintaining the original slope displacement analysis, the thermal image can also be used to analyze the temperature difference and potentially unstable areas behind the slope for future planning and tracking. In this study, after the development of the field instrument, the scale model test will be used to evaluate the instrument. The slope behind the Central University is then applied as a field example. By setting up a monitoring system on site, the slope displacement analysis is performed from the multi-phase images, and the monitoring process of the overall system is realized through the three-dimensional information model of post-processing. According to the experimental results, the accuracy can reach within 3 cm. The results can provide more effective decision data for ii slope disaster warning, prevention, and post-disaster planning.
