CAJ | 학술논문

提出了一种基于分形维数和FCM聚类的SAR图像无监督变化检测的算法。首先用非下采样Con-tourlet变换（NSCT），对两时相图像进行分解，然后求出其分形维数图，构造差异图，再由FCM聚类得到变化区域和非变化区域。并对计算分形维的滑动窗口大小的选择进行了研究。与现有的基于分形维数的图像变化检测算法进行对比，实验证实，本文算法不仅对斑点噪声不敏感，并且提高了变化检测的精确度。
제출료일충기우분형유수화FCM취류적SAR도상무감독변화검측적산법。수선용비하채양Con-tourlet변환（NSCT），대량시상도상진행분해，연후구출기분형유수도，구조차이도，재유FCM취류득도변화구역화비변화구역。병대계산분형유적활동창구대소적선택진행료연구。여현유적기우분형유수적도상변화검측산법진행대비，실험증실，본문산법불부대반점조성불민감，병차제고료변화검측적정학도。
This paper has proposed a unsupervised SAR image change detection algorithms based on fractal di-mension and FCM clustering.First, the two-phase images are decomposed by nonsubsampled Contourlet transform (NSCT) and then compute its fractal dimension map and get difference map. At last, the change map is obtained by the FCM clustering. And the size of sliding window of calculating the fractal dimension is studied. Experiments con-firme that, compared with the traditional image change detection algorithm based on fractal dimension (GFCD), the proposed algorithm is not only sensitive to the speckle noise but also improves the change detection accuracy.