高分辨率遥感影像的形态学边缘检测算法

MORPHOLOGY EDGE DETECTION ALGORITHM FOR REMOTE SENSING IMAGES WITH HIGH RESOLUTION

结合高空间分辨率遥感影像的特点,利用数学形态学方法,选取具有不同尺度和包含全部方向的结构元素,设计一种多尺度全方位形态学边缘检测算法,并采用峰值信噪比方法代替均值以自适应地确定尺度权重。实验结果表明:与传统Canny算子和结构元素尺度形状单一的形态学边缘检测算法相比,本算法解决了噪声抑制和精细边缘提取的矛盾,检测出的图像边缘清晰,而且抗噪性能强。边缘检测结果可用于遥感影像地物几何特征和纹理特征提取。

Based on the characteristics of remote sensing images with high spatial resolution and mathematical morphology （MM） methods, a multi-scale and omnidirectional MM edge detection algorithm were developed by using multi-scale omnidirectional structure elements. In this algorithm,the peak signal-to-noise ratio （PSNR） technique was adopted to replace the method with fixed mean for acquiring the self-adaptive scale weights. Experimental results suggest that ： the proposed algorithms, comparing with the traditional Canny edge detection operator and MM edge detection algorithm with simplex scale figure structure elements, can successfully resolve the contradiction between noise suppression and fine edge extraction excellently, the image edge detected is distinct and its anti-noise performance is strong. The detected edge information can be used in the extraction of geometrical and texture features of remote sensing image landmarks.