基于均值漂移的高分辨率影像多尺度分割(英文)

VHR Imagery Multi-Resolution Segmentation Based on Mean Shift

高空间分辨率遥感影像在许多领域均有应用。由于遥感影像数据量大且内容复杂,目前少有针对这种影像的有效分割方法。引入一种快速、稳健的多尺度分割方法——均值漂移,该方法是一种通过简单迭代快速自适应上升的模式搜索法。基于均值漂移算法的分割方法,并充分利用光谱特征与空间特征,通过具有一定物理意义的参数控制分割精度,与目前商用软件eCognition提出的分割算法相比,同样达到与视觉分割一致的效果,并且速度更快。

The very high spatial resolution satellite images have been applied into many fields. However,researches on segmentation of such kinds of images are rather insufficient partly due to the complexity and large size of the images. In this study,a fast and robust multi-resolution segmentation approach is proposed based on mean shift,a simple iterative procedure that shifts each data point to the average of data points in its neighborhood,which has been proven to be a mode-seeking process on a surface constructed with a ＂shadow＂ kernel. In the presented algorithm,not only the color features,but also the space relationship of each pixel is considered in multiple scales. Compared with the segmentation approach of a commercial software eCognition,the proposed one is a bit faster when applied to the Quickbird images and has a satisfactory segmentation result like eCognition.