像元与对象特征融合的高分辨率遥感影像道路中心线提取

提出了一种融合像元-多尺度对象级特征的高分辨率遥感影像道路中心线提取方法。首先在像素级上提取影像的纹理和形状结构特征,在构建的多尺度分割集影像上提取对象的区域光谱特征。然后,将像元级特征与多尺度对象特征进行决策级融合,完成道路网的粗提取。最后,结合本文所提出的非道路区域自动去除算法和张量投票算法,实现道路中心线的精提取。不同场景、不同分辨率数据下开展的试验结果表明,该方法可有效改善传统道路提取方法易产生的"盐噪声"和非道路地物粘连现象。

Fusion of Pixel-based and Object-based Features for Road Centerl ine Extraction from High-resolution Satellite Imagery

A novel approach for road centerline extraction from high spatial resolution satellite imagery is proposed by fusing both pixel-based and object-based features.Firstly,texture and shape features are extracted at the pixel level,and spectral features are extracted at the object level based on multi-scale image segmentation maps.Then,extracted multiple features are utilized in the fusion framework of Dempster-Shafer evidence theory to roughly identify the road network regions.Finally,an automatic noise removing algorithm combined with the tensor voting strategy is presented to accurately extract the road centerline.Experimental results using high-resolution satellite imageries with different scenes and spatial resolutions showed that the proposed approach compared favorably with the traditional methods, particularly in the aspect of eliminating the salt noise and conglutination phenomenon.