本論文特以數值高程模型(DEM)分析地形，利用電腦程式自動化萃取影像中的線形資訊，並進一步分析線形方向分佈，以期提供地質學者分析應用。 數值高程模型利用光達或影像對，計算了各個位置的高度值，利用此資訊可進一步計算其坡度。本研究流程首先對原始影像使用高斯金字塔降低雜訊與小線形的影響，接著利用透空度分析(Openness)，計算範圍圓內的八個天頂角以及天底角，再經由期望值最大(EM)演算法計算門檻值萃取出山脊線及山谷線。 最後在線段追蹤以及方向統計，將萃取出的山脊部份，使用細線化將每個線段化簡成一個像元寬度，使用Lee & Jyrkenvich所提出的線段追蹤演算法進行線段追蹤，再進行每個點的方向統計，以分析山脈走向。最後將影像分成許多小區塊，並分別統計方向，分析不同區塊的山脈走向。 我們使用台灣西南地區的數值高程模型作為測試資料進行實驗，並於高通濾波與人工結果比較驗證 In this paper we analyzed the topography from Digital Elevation Model (DEM) and designed a computer aided algorithm to automatically extract the lineament information in the image. In addition, we also analyzed the distribution of direction of lineament. The extracted information offers useful data to geological analysis and their applications. DEM providing the elevation value of each coordinate is calculated from LIDAR or the image pairs. It can also be used to calculate the slope. The proposed procedure has four steps. Firstly, it uses the Gaussian Pyramid to reduce the influence of noise and short lineament from the original image. Secondly, Openness is adapted to calculate the zenith and nadir angle of eight azimuth in circular range. And then the data is thresholded to extracts the mountain ridge and valley line via EM algorithms. Finally, in line tracing and gathering directional statistics, we use thining rule to make the extracted ridge to one pixel width, then track the line with algorithm proposed by Lee & Jyrkenvich. In direction distribution analysis we calculate the direction of each pixel on the extracted line and draw the histogram for the whole area. We also apply the same procedure to the small blocks divided from the image to have localized analysis. We use the DEM of the southwest of Taiwan as the test data to carry on the experiment, and to verify in the high pass filtered and artificial result.
