Abstract: Taking into account the shape of a road and the characteristics of gray value variation on roads’ surfaces，a new strategy is given for high resolution remotely sensed images.By comparing the consistency of gray value and one dimensional texture in different directions，the most consistent direction at every pixel is found.The responses of a group of Gabor filters together with the variance indicating the consistency in this direction consist of the feature vector.Starting from a pixel a short line segment along the most consistent direction at this point can be drawn，and all the pixels lying on this short line are defined as the neighbor of the point.With the aid of the newly defined neighbor an improved spatial FCM segmentation model is obtained.Experiments demonstrate that the proposed method can extract roads from other targets in high resolution images more efficiently than some commonly used algorithms.

Key words: road extraction, image segmentation, spatial fuzzy C-means model, texture analysis

摘要： 提出了一种从遥感影像中分割出道路的新策略，其最大的创新之处在于充分考虑了道路的形状特点及沿道路方向的灰度分布特性。比较多个方向上纹理和灰度的一致性，得出一致性最优方向，此方向上的Gabor滤波响应和作为衡量一致性指标的均方差值共同构成本方法的特征矢量。从每一像素出发，沿该点的一致性最优方向绘制一条短线，用短线对应的区域代替现有空间信息FCM模型中的邻域以获取改进的分割模型。实验表明：这一建立在新的特征矢量和分割模型基础上的方法可以更为有效地从高分辨率遥感影像中分割出复杂的道路目标。

关键词: 道路提取, 图像分割, 空间信息模糊C均值（FCM）模型, 纹理分析