基于多重分形的遥感图像融合检测方法  

Remote Sensing Image Target Fusion Detection Method Based on Multifractal

作　　者：田微晴 程慧华[1] 刘琼俐[1] 周鹏[1] 李津[1] 

机构地区：[1]武汉军械士官学校,武汉430075

出　　处：《四川兵工学报》2015年第3期135-137,共3页Journal of Sichuan Ordnance

摘　　要：多重分形通过奇异指数和多重分形谱分别从局部和全局描述了图像的特征。研究了多重分形的定义及其计算方法,分别提取了SAR图像和可见光图像的4个多重分形谱特征。为了提高聚类分割精度,将图像的灰度特征和4个多重分形谱特征组成特征矢量,作为模糊C均值聚类的输入,对SAR图像和可见光图像进行了分割。多重分形谱特征从全局反映了图像的全局特征,为此,选取两幅图像的多重分形谱特征为依据,对可见光图像和SAR分割图像进行"折中"融合,减小检测的虚警和漏警率。实验结果表明,采用该方法能结合两幅图像各自的优势,有效提高检测的精度。Multifractal singularity and the multifractal spectrum describe the characteristics of the image from the local and global respectively. This paper first described the multifractal and how to calculate multifractal spectrum based on different measures. Four multifractal spectrum characteristics of the SAR and visible images were extracted. To improve the accuracy of segmentation,the paper took the gray features and the four multifractal spectrum characteristics as the input of fuzzy C-Means clustering,and completed the segmentation of SAR image and visible image. The multifractal spectrum reflects the global characteristics of image,in this case the multifractal spectrums of the two images were used to finish the image fusion. The results show that the method of the paper has fully combined the advantages of the two images to improve the detection accuracy.

关 键 词：分形 多重分形 多重分形谱 分割 融合检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]