抗旋转和缩放的SAR与可见光图像自动配准算法

摘要

针对合成孔径雷达与可见光图像在大角度旋转和大比例缩放情况下的高精度自动配准问题,提出了一种尺度和旋转不变的SAR(Synthetie Aperture Radar)和可见光图像自动配准算法.算法以SIFT(Scale Invariant Feature Transform)算法为基础,首先通过增强Frost滤波和自适应直方图均衡增强SAR和可见光图像的共性,使其显著提高能够提取出足够多的特征点数目,然后再通过特征描述方法、相似性度量方法、点匹配方法、特征点聚类方法和误匹配点剔除方法等方面对原始SIFT方法进行改进,有效地提高其在多源图像、强噪声、复杂成像条件下的特征提取和匹配性能,最后通过最小二乘法和相似变换模型实现SAR和可见光图像的精确配准.试验表明该算法对图像尺度和角度变化具有良好的适用性,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Harris算法,具有良好的工程应用前景.%Focus on the image regiatration problems for SAR ( Synthetic Aperture Radar) and optical images on condition that existing big scale and angle change, a novel image registration algorithm based on improved SIFT( Scale Invariant Feature Transform) is proposed. First, the enhanced Frost Filter and adaptive local histogram equalization is adopted to improve the common property of SAR and optical images to get enough SIFT points and high represent rate. Then multi-main-direction assignment, normalized cross-correlation measure, clustering analysis, and full hypothesis-proof-test methods are introduced into the algorithm lo improve the key point detection, point-pair matching and out-point-pair eliminating performance. Finally, based on the reserved points, the images are registered automatically by using similarity transformation model and least squares method. The experiment results approve that the propoaed method is better than SIFT, Harris and other popular state-of-art algorithms both in correct matching probability and positioning accuracy, and would have a good prospect in engineering application.