结合路径形态学的高分遥感影像道路提取方法

Road Extraction Method Based on Path Morphology for High Resolution Remote Sensing Imagery

随着光学遥感影像空间分辨率不断提高,车辆、阴影、标识牌等地物以及油返砂对道路提取构成更严重的干扰,路径形态学的道路提取方法虽可以有效地剔除这些干扰,但仍然存在道路粘连和阴影遮挡问题。针对上述两问题,首先采用Otsu方法对影像进行分割得到初始道路,其次应用路径开运算得到改进道路提取结果,然后提出形态学与路径形态学结合方法以削减道路粘连,最后基于改进圆形模板跟踪模型对断裂点进行跟踪,以解决阴影遮挡造成的道路断裂问题。通过多组影像的实验对比分析表明,该方法可以明显改善高分遥感影像道路提取效果。

With the continuous improvement of spatial resolution of optical remote sensing image,the ground objects such as vehicles,shadows,signage,and oil return sand pose more serious interference to road extraction.Although the road extraction method of path morphology can effectively eliminate the interference,there are still problems of road adhesion and shadow blocking.This paper put forward an improved road extraction method.Firstly,the Otsu algorithm was used to segment original images to obtain initial results of road extraction.Secondly,the path opening was applied to obtain improved results.Thirdly,a morphology method combined with path morphology was presented to eliminate the adhering non-road.Finally,in order to resolve the problem of the road fracture caused by shadow masking,a tracking model based on improved circular template was presented to track breakpoints.Comparative analysis of experiments of multi-group images showed that the presented method could significantly improve the effect of road extraction from high resolution optical satellite images.