一种新的遥感影像边缘检测方法

A New Edge Detection Method of Remote Sensing Image

提出一种针对于遥感影像的边缘检测方法。该方法首先利用多个尺寸的邻域结构对图像进行平滑,然后根据图像灰度进行对象云化处理,构建对象云。通过云运算生成边界云并构建边缘模糊特征平面,在条件概率和模糊划分熵的基础上,通过最大模糊熵原则确定最优阈值,对图像模糊边界进行提取。试验结果表明,该算法能保留大量低灰度信息,并有效地去除了次要边缘对主边缘的干扰。

A new edge detection algorithm of remote sensing image is proposed. First, smoothing the image by different masks, we can construct the object-clouds based on gray-level of images by region growth algorithm. Second, constructing boundary-clouds and fuzzy property facet by Boolean calculation between two or more clouds. Third, getting the optimal threshold through maximizing the entropy of fuzzy partition based on conditional probabilities and entropy of fuzzy partition. At last, we can get the edge of aerial image by fuzzy property facet and optimal threshold. We conduct various performance to show that plenty of low-level gray can be preserved and the disturb of small edges to main edge is suppressed.