优化区域生长的高分辨率影像道路精确提取

鉴于区域生长存在需提供种子点及易过生长的缺陷,改进了区域生长算法对高分辨率遥感影像进行道路精确提取。首先利用格网对原始影像进行划分,将在道路灰度范围内的外边缘格网以同质性指标依次排序作为候选种子点;将满足道路连续性和一致性要求的路径格网起始点作为种子点;并通过道路指数加以验证得到最终种子点,以此避免盲目选择种子点的不足。其次在引入梯度值进行道路生长后,以二值化的彩色分量与生长结果进行信息融合方式来优化算法,达到缓解过生长的目的。实验结果表明算法是合理有效的,能够较为精确地提取出道路区。

Precise Road Extraction of High-resolution Imagery with Optimized Region Growing

In view of the region growing with defects of being provided seed points and overgrowth. In this paper, the regional growth algorithm is improved to extract high-resolution remote sensing images accurately. Firstly, the original image is divided by the grid, and the outer edge grids in the gray scale of the road are sorted by the homogeneity index as the candidate seed points. The starting point of the path grid can meet the requirements of continuity and consistency of the road is the seed points, and it is verified by road index to get the final seed points, so as to avoid shortage of blind selection of seed points. Secondly, after the introduction of the gradient value for road growth, the binarized color component and the growth result are used to optimize the algorithm to achieve the purpose of alleviating the growth. The experimental results show that the algorithms are reasonable and effective. And the algorithms can extract the road area more accurately.