Abstract: An improved region growing method for road extraction in the remote sense images is proposed. K-means clustering algorithm is used to separate the road region from the non-road region of remote sensing images, gaining judgment conditions for region growing and getting the threshold value according to the image feature. Nine road junction models are designed according to the road characteristics. Based on the proposed road models, the region growing method is improved to extract the road in a way of the road intersection automatic growth. The mathematical morphology is utilized to optimize road region. The experimental results show that the extracted road region is more complete than that of literature method.

Key words: remote sensing images, spectral information, road extraction, region growing, road junction model

摘要： 提出一种改进区域生长法的遥感影像中道路提取方法。对遥感影像进行[K]均值聚类，实现道路区域和非道路区域的初步分离，并获取区域生长的基准值，按照图像特征计算出区域生长的阈值。依据对道路特性的分析，设计了9个道路路口模型。根据设计的道路路口模型，对区域生长法进行了改进，使得道路的提取按照道路路口模型自动增长。最后通过数学形态学的手段对道路进行优化。实验结果表明使用提出方法所提取道路区域更加完整。

关键词: 遥感影像, 光谱信息, 道路提取, 区域生长, 路口模型