顾及梯度和彩色信息的高分辨率影像道路分割

Road Segmentation of High-spatial Resolution Remote Sensing Images by Considering Gradient and Color Information

针对高分辨率影像中道路灰度值分布不均,“同谱异物”现象难以用简单阈值分割,以及对不同传感器、不同分辨率影像提取效果不一样等问题,提出了一种顾及空间梯度信息和彩色信息的道路阈值分割方法。利用高分辨率影像的多种特征,将梯度值和灰度值进行联立得到相关性连接因子后进行分割;同时将原始道路影像转换到HSI彩色空间分割出道路灰度一致性的区域,结合梯度值对分割结果进行边界验证;最后将验证结果和相关性连接因子分割结果进行融合,剔除粘连的非道路信息,过滤非道路信息提取出精确道路。采用不同传感器、不同分辨率的影像进行实验,结果表明算法可以较好地解决上述问题且适应性强,通过与现有方法结果对比,验证了所提算法在精度上的优越性。

A road threshold segmentation method that taking into account spatial gradient information and color information was presented in order to solve the problem of uneven distribution of road gray values in high resolution images and the difficulty of segmentation of the different things with the same spectrum with simple thresholds,and it has different effects on the extraction of high-resolution road images from different sensors and different resolutions.The gradient value and gray value are combined to get the correlation connection factor,and the original road image is transformed into HSI color space to segment the gray consistency region of the road,and the segmentation results are verified by gradient value.Finally,the validation results and the segmentation results of correlation connectivity factors are fused to remove the non-road information,and filter the non-road information to extract the accurate road.Experiments with different sensors and different resolutions show that the algorithm can solve the above problems well and adaptability.Compared with the existing methods,the superiority of the proposed algorithm is verified.