基于模糊控制的ASIFT图像特征优化算法  

Affine SIFT Feature Optimization Algorithm Based on Fuzzy Control

作　　者：蔡国榕[1,2,3] 李绍滋[1,2] 陈水利[3] 吴云东[3] 苏松志[1,2] 

机构地区：[1]厦门大学智能科学与技术系,福建厦门361005 [2]福建省仿脑重点实验室(厦门大学),福建厦门361005 [3]集美大学理学院,福建厦门361021

出　　处：《模糊系统与数学》2012年第5期147-153,共7页Fuzzy Systems and Mathematics

基　　金：国家自然科学基金资助项目(61202143;61103052);高等学校博士学科点基金资助项目(20090121110032);深圳市科技项目(JC200903180630A;ZYB200907110169A);福建省产学重大项目(2011H6020);福建省自然科学基金资助项目(2011J01013);福建省教育厅项目(JA10196)

摘　　要：ASIFT算法是图像特征匹配的有效工具,具有很强的仿射不变性。在ASIFT算法的基础上,利用Nelder-Mead单纯形方法优化采样点,并通过模糊控制策略实现单纯形参数的自适应调整。针对一组低空遥感图像的实验结果表明,本文方法保持了ASIFT算法对仿射变换的鲁棒性,并且能获得比ASIFT和SIFT算法更多的特征匹配对。Affine SIFT is an efficient tool for image matching, which has been proven to be invariant to affine distortion. To acquire more matches between reference image and input image, this paper used Nelder-Mead Simplex to optimize the best affine transformation based on transforms acquired by ASIFT. Moreover, fuzzy control has been employed to construct an adaptive parameter tuning strategy of Simplex. Experiments conducted remote sensing images show that the proposed algorithm is also invariant to affine distortion. Furthermore, the proposed method achieved promising results since we acquired more correct matches than ASIFT and SIFT.

关 键 词：SIFT特征 仿射变换 Nelder-Mead单纯形法 模糊控制 

分 类 号：O159[理学—数学]