基于感知编组的SAR图像建筑物检测算法

针对合成孔径雷达(SAR)图像自动解译的需求,研究了高分辨率的SAR图像建筑物检测技术。由于人造目标按照某种方式有组织的构成,而机器视觉可以模拟人类视觉的感知编组能力获取图像的相关组织和结构,并可以根据特定的准则将提取到的图像特征编组成为更高层的结构,利用该特性,将感知编组方法用于检测建筑物,通过边缘检测和霍夫变换提取直线段基元,结合感知编组和建筑区在SAR图像上表现的亮度特征,提出了一种平行适应度函数辅助建筑物检测。基于对实测SAR图像的实验结果表明,该检测算法可靠,定位准确,并能有效地降低虚警率。

Building Detection Algorithm of SAR Images Based on Perceptual Organization

In view of the application requirements of Synthetic Aperture Radar （SAR）image automatic in-terpretation,this paper have researched the building detection technology in high resolution SAR Image. As man-made objects are organized by some certain rules,machine vision can simulate perceptual organi-zation of human vision to obtain the man-made organization and structure of the image,we can organize the image feature have been extracted to be high-rise structure by some certain rules,so we will use perceptual organization to detect building.In this paper,we extract primitive line through edge detection and Hough Transform,and then combine the perceptual organization theory and the brightness feature of the building area in SAR image,introduce a parallel fitness function to assist building detection.Our experiments result based on real SAR image show that the detection algorithm is reliable,the position of buildings detected by the method is accuracy,and it can effectively reduce the false alarm rate.