一种高分辩遥感影像多尺度分割新算法

Multi-scale Image Segmentation of High Spatial Resolution Remote Sensing Image Based on Mean-shift Procedure

提出了一种稳健的高分辨率遥感图像多尺度分割算法。该方法基于自适应Mean-shift聚类过程,不需要假定数据分布类型,也不需要指定类别的数目,自动化程度较高,并针对高分辨遥感图像所体现出来的地物的多种信息特征,结合多种特征进行了区域合并。采用分辨率为1 m的IKONOS卫星影像进行分割试验。结果表明,该算法不仅能充分利用高分辨率遥感图像中地物的信息特征获得良好的分割效果,而且自适应程度高,抗噪能力强,精度也能满足要求,是一种稳健的自动分割方法。

A steady hierarchical segmentation algorithm for high-resolution remote sensing image is proposed in this paper. The method is based on adaptive Mean-shift clustering; it does not need assumption of data distribution type and designation of cluster number. Regions are merged with multi-feature of terrain and object in high-resolution remote sensing image. Segmentation experiments were processed with lm resolution IKONOS satellite image. The test results show that the presented method is adaptive and robust to noise, it can make the most of multi-feature of terrain and object in high-resolution remote sensing image, the precision is satisfied. So it is a steady automatic segmentation algorithm.