基于直觉模糊C核均值聚类算法的合成孔径雷达图像变化检测

Synthetic Aperture Radar Image Change Detection Based on Intuitionistic Fuzzy C-Core Mean Clustering Algorithm

针对合成孔径雷达图像中存在椒盐噪声影响变化检测精度的问题,在变化检测步骤中应用一种改进的直觉模糊C核均值聚类算法;首先运用代数运算方法——差值法、比值法、图像回归法,构建3幅光谱变化差异的图像,将3幅差异图像组成列向量;然后使用主成分分析算法对差异图像构成的列向量进行特征提取;最后采用直觉模糊C核均值聚类算法获取影像变化图。实验结果表明:所用算法可以减小椒盐噪声对合成孔径雷达影像变化检测的影响,而且该算法能更好地保留特征信息,提高变化检测的精度。

Synthetic aperture radar images are characterized by salt&pepper noise that affects change detection accuracy.To solve this problem,an improved method based on intuitionistic fuzzy C-core mean clustering is applied to synthetic aperture radar image change detection.First,we use algebraic operation methods,such as the difference,ratio,and image regression methods,to construct three types of spectral variation difference images,and three different images are grouped to column vectors.Second,the principal component analysis algorithm is used to extract features from column vectors of the difference images.Finally,the change map is obtained using intuitionistic fuzzy C-core mean clustering.Experimental results show that the proposed method can reduce the influence of salt&pepper noise in synthetic aperture radar images,retain feature information,and improve the accuracy of change detection in the image.