基于阈值标记的分水岭算法道路提取

Road Extraction Based on Threshold Labeling Watershed Algorithm

摘要: 分水岭算法易产生过分割现象，本文提出一种基于阈值标记的分水岭分割方法，能更为精准提取遥感图像中的道路信息。该方法对遥感图像进行中值滤波及直方图均衡化预处理；经多尺度形态学梯度计算出最大梯度值、进行低通滤波，然后使用二维最大熵求阈值、完成扩展的极小值变换，通过这些标记对梯度幅度图像加以修正；最后实现标记梯度图像的分水岭变换，利用形状特征提取图像中的道路信息。本文使用MATLAB仿真软件对银川某地的遥感图像进行道路提取，实验结果表明：该算法可以提取出较为精确、清晰的道路信息。

Abstract: Aiming at the phenomenon that the watershed algorithm is easy to produce over segmentation, a watershed segmentation method based on threshold marking is proposed to extract road information in remote sensing image. Firstly, the median filtering and histogram equalization of remote sensing images are pre-processed, and then the pre-processed image is processed to obtain the multi-scale morphological gradient, and the maximum morphological gradient is obtained by low pass filtering. Then threshold is calculated by the 2D maximum entropy, extending minima transform with the income threshold, using these markers to modifier gradient magnitude image; finally, the marked gradient image was carried out by watershed transform using shape feature to extract the road information in the image. MATLAB is used to extract the road from a remote sensing image of Yinchuan. The experimental results show that the algorithm can extract the road information in the image very well.