结合均值漂移分割与聚类分析的遥感影像变化检测

针对传统遥感影像变化检测方法对前后期影像数据质量要求高、适应范围窄、检测精度较低等问题，本文在引入异常点检测思想的基础上，提出了一种结合均值漂移分割与聚类分析的遥感影像变化检测方法。首先将前期地理国情矢量数据与待监测的遥感影像进行地理配准；然后在地理国情矢量数据的基础上对遥感影像作均值漂移算法二次细分割，获得的矢量图斑继承了原有父级类属性，并对同一父级类的影像图斑进行光谱、空间、纹理、指数等特征提取；最后采用高斯混合模型与最大期望值算法聚类获得多个类别，在父级类下找出异常类别的图斑。通过试验对比分析，表明了本文方法的有效性和可靠性，为遥感影像变化检测提供了新思路。

Remote Sensing Imageries Change Detection Combined with Mean-shift Segmentation and Cluster Analysis

In order to solve the problem that the traditional remote sensing imageries change detection method has high quality requirements ,low adaptability range and low detection precision , on the basis of introducing the idea of outlier detection , a remote sensing imageries change detection method combined with mean-shift and cluster analysis is proposed in this paper .First,early stage geographic situation vector is used to register with the RS image ,and the multi-scale subdivision of the remote sensing imageries is done . The resulting small image inherits the original pattern category attribute , and then extracts the spectral , geometric, texture, and exponential features of the image .And then clustering the extracted multi -feature based on GMM-EM.According to the type of clustering,the change pattern is determined by comparing with the original vector pattern .The experimental results show that the method is effective and reliable ,which provides a new idea for remote sensing image change detection .