Gauss比较函数与复杂遥感图像中线状目标的无偏自动检测

A Gaussian Comparison Function and Unbiased Detection of Linear Structures from Remote Sensing Images

建立基于Gauss尺度空间的比较函数,对遥感图像中的道路结构进行特征描述、分离和定位.在此基础上,结合道路特征的图形和图像特征,提出了基于局部方向能量的线状目标检测算法,并根据道路的拓扑特征和几何特征进行假设验证、编组、融合,提取有效的道路线特征,应用于城市遥感图像中不同宽度和材质的主干道路和小路的提取.该算法计算复杂度小,在阴影遮挡和道路影像不明显的情况下对道路线特征具有良好的分辨能力.对Gauss比较函数的定位和抗零点漂移性能也进行了详细的分析.

Based on the Gaussian scale-space, a Gaussian comparison function is presented for separating the adjacent road features in aerial image. Combining the geometric and radiometric features of the roads, the curvilinear structures are extracted based on local directional energy in continuous scale-space. Valid road curvilinear features are verified, grouped and extracted by both topologic and geometric methods. The above algorithm is applied to extracting the road features of both rural road and urban highways from urban aerial images. The algorithm can efficiently extract non-salient road and shadowed road, and can significantly reduce the computational complexity in the line tracing. Some discussions on the zero drift of the Gaussian comparison function are also presented.