基于VGA聚类的遥感影像道路提取

Remote Sensing Image Road Extraction Based on VGA Clustering

针对地表情况及地物类型不确定的高分辨率遥感影像道路提取问题,提出了一种结合可变串长遗传算法(VGA)和模糊C聚类的道路目标提取方法。首先,通过可变串长给出实际影像可能的聚类;然后,以模糊相似性测度为基础构造适应度函数,采用遗传算法自动确定适合的聚类数及最优模糊隶属度矩阵,并对影像进行聚类分割;最后,针对分割结果进行道路及中心线的仿真提取。提出的方法不仅有效解决了遥感影像道路提取时因聚类数目及聚类中心无法预先获取而带来的问题,而且避免聚类分割陷入局部最优,实现了全局搜索,缩减了目标函数收敛时间。通过对Ikonos、Quick Bird和航拍等高分辨率遥感影像进行仿真,结果表明提出的方法可以有效且高精度地提取出影像中的道路目标,为遥感影像的实际应用奠定了基础。

In order to solve the problem of road extraction of high resolution remote sensing images with uncertain surface and ground object types, the road object extraction method based on variable string length genetic algorithm （VGA） and fuzzy C clustering is proposed. Firstly, the possible clustering number of remote sensing images was given by variable string length. Secondly, the fitness function was defined based on the fuzzy similarity measure, the genetic algorithm was used to automatically determine appropriate number of clusters and the optimal fuzzy membership matrix, and the remote sensing image is clustered and segmented. Finally, the results of road and centerline were extracted according to the result of segmentation. The proposed method not only effectively solves the problem that the number of clusters and the clustering center can not be pre-acquired for road extraction of remote sensing images, but also avoids the clustering segmentation falling into the local optimal solution, achieves global search and reduces the convergence time of the objective function. Simulation has been conducted on high resolution color Ikonos, QuickBird and Aerial images. The results show that the proposed method can extract the road objects from images effectively and accurately, which lays the foundation for the practical application of high resolution remote sensing images.