北京建筑大学学报

以中国江西省南昌市南昌县2015年和2017年的高分一号16 m分辨率遥感影像为例,进行协同分割变化检测.协同分割变化检测算法引入了计算机视觉中的多图像协同分割思想,利用变化强度图作为引导,对图像进行目标和背景的分割.算法通过建立网络流图,将图像分割的问题转化为能量函数最小化问题.该算法利用基于增广路径的Dinic算法将能量函数最小化,在求得图像的最小割的同时得到最终的分割结果.分割结果的总体精度约为0. 834,kappa系数约为0. 663,可见面向高分一号遥感影像,协同分割的变化检测方法可以较为准确地提取出变化对象,实现大范围的变化检测.

Taking the 16-meter resolution remote sensing image of Gaoguan No. 1 in Nanchang County,Jiangxi Province,China as an example,the cosegmentation change detection was carried out. In this paper,the cosegmentation algorithm is used to introduce the idea of multi-image cosegmentation in computer vision. The change intensity map is used as the guide to segment the target and the background. By establishing the network flow graph,the problem of image segmentation is transformed into the minimum energy function. The problem is to map the image into a network flow graph,and use the Dinic algorithm based on the augmented path to obtain the minimum cut of the graph to minimize the energy function and obtain the final segmentation result. The overall accuracy of the segmentation result is about 0. 834,and the kappa coefficient is about 0. 663. It can be seen that the high resolution remote sensing image is used for the high resolution remote sensing image,and the change detection method of the cosegmentation can extract the change object more accurately and realize the wide range change detection.