基于多特征融合的FCM面向对象变化检测

FCM Object-oriented Change Detection Based on Fusion of Multi-Feature

光谱特征和纹理特征的差异是遥感影像进行变化检测重要依据,多特征的融合能够一定程度弥补单一特征的不足。模糊C聚类(FCM)能够根据多个特征进行聚类,通常在FCM聚类中各个特征都是等权的,这样并不能体现各个特征的权重差异。论文尝试改进FCM算法进行多特征融合变化检测,首先利用分割后影像的光谱特征、纹理特征分别进行变化检测,然后将不同的特征检测结果进行直接加权和FCM加权,结果表明论文方法比传统的单一特征和多特征融合方法变化检测有一定的优势。

The difference between spectral features and texture features is an important basis for remote sensing image change detection.Multi-feature fusion can compensate for the deficiency of single feature to a certain extent.FCM can cluster according to multiple features.Generally,each feature is equal weight in FCM clustering,which cannot reflect the weight difference of each fea⁃ture.This paper attempts to improve the FCM algorithm for multi-feature fusion change detection.Firstly,the spectral and texture features of the segmented image are used to detect the change,and then different feature detection results are directly and FCM weighted.The results show that the proposed method is better than the traditional single feature and multi-feature fusion method for change detection.