（文献）基于改进SIFT的SAR图像与可见光图像配准.pdfVIP

Dec2010 13 V o l28, No. 6 A erospace Contro l SIFT SAR , 430074 针对 SAR 图像与可见光图像的自动配准问题, 提出了一种基于尺度不 变特征( S IFT 的 SAR 与可见光图像配准算法算法首 对 SIFT 主方向的检测 进行了优化, 利用特征点邻域内边缘局部极大值点的梯度方向作为特征点的方 向, 并以去均值归一化互相关系数为相似性度量进行特征点对匹配, 然后通过随 机抽样一致性算法( RAN SAC 剔除误匹配点对, 最后利用剩余的特征点对实现 SAR 与可见光图像的自动配准实验结果表明, 本算法对不同分辨率图像和不 同旋转角度图像具有较好的适用性, 在正确匹配点的比率和定位精度方面都优 于原始 SIFT 算法和H arris算法 改进 SIFT; SAR 图像; 可见光图像; 特征点匹配; 图像配准 : T P751. 1 : A : 2010 Registration A lgorithm for SAR and Optical Im ages Based on Improved SIFT YANG X uem ei GONG Junb in WANG P eng T IAN J inw en IPRA I o fHU ST State K ey Laboratory for M ultispectral Inform ation P rocessing T echno log ies, W uhan 430074, Ch ina Abstract Focusing on the automatic mi age reg istrationproblem of SAR and op tical images, a new algo rithm based on SIFT is presented in thispaper. Firstly, the detection of SIFT dom inant direction is op ti m ized, in which the gradientdirections of localmaxima edgepoints are used as thedirections of key point. Secondly, the normalized crosscorrelated algorithm is used in the f eatures matching. Thirdly, the RAN SAC algorithm is app lied to removefalsematching points. Finally, according to the correctmatching p oints, the mi ages are registered automatically. In the experiments, theadap tability of thealgorithm is an alyzedfor the images of dfif erent resolutions and dfif erent rotating angles. The exp erimentresults