基于知识的高分辨率影像长城目标提取方法研究

Study on extracting great wall object based on knowledge from high-resolution image

通过对高分辨率影像上长城特征的分析,本文提出了“在山地地貌条件下长城沿山脊线分布”的推理假设,基于该假设,进一步提出了以基于梯度方向滤波的边缘提取、基于DEM山脊特征的长城边缘线段判定以及基于区域特征的线状区域分割相结合的长城提取算法。

According to the Great Wall targets on the images and the former research, the characters of the Great Wall on image are summarized firstly. Then the assumption that the Great Wall distributes along the ridge in the mountain areas is described for the design of the extraction algorithm. The whole algorithm for the Great Wall extraction is designed as this: Edge extraction based on gradient direction filter, the Great Wall edge linear section judgment based on the ridge information extracted from DEM, linear region segmentation based on the region attributes. This algorithm is proved efficient by the extraction experiment using 1m DOM.