论文部分内容阅读
图像配准是遥感图像处理中的基本问题.本文针对多源多时相遥感影像的特点,提出了一种基于自适应尺度的渐进配准方法,在从粗到细的迭代配准过程中,可以通过上一次配准结果的几何定位误差来确定本次匹配的尺度,并按该尺度提取特征角点和特征邻域进行匹配,与常规金字塔渐进配准方法相比,减少了匹配次数,提高了配准效率.另外,特征提取和匹配过程中提出一种基于Harris-Laplace算法和相位相关算法的遥感影像配准算法,利用Harris-Laplace角点代替原始图像,能够综合区域和特征的优点,对亚像元偏移、旋转、尺度变化具有不变性,同时对对比度和灰度的变化不敏感,具有很强的抗噪性.在特征检测和匹配的过程中采用限定搜索区域、抽稀角点等多种优化策略来提高算法的性能.实验证明,算法具有很好的精度,对几何攻击具有很好的鲁棒性,该算法已经应用于CBERS-02B星3级数据的批量自动化生产,具有很好的应用效果.
Image registration is a basic problem in remote sensing image processing.Aiming at the characteristics of multi-source and multi-temporal remote sensing images, a progressive registration method based on adaptive scaling is proposed. In the coarse to fine iterative registration process, Through the geometric registration error of the previous registration result, the scale of this match is determined, and the feature corner and feature neighborhood are extracted according to the scale to match. Compared with the conventional pyramid progressive registration method, the matching times are reduced and the matching times are increased Registration efficiency.In addition, a remote sensing image registration algorithm based on Harris-Laplace algorithm and phase correlation algorithm is proposed in the process of feature extraction and matching. Harris-Laplace corner can be used instead of original image to synthesize the advantages of region and feature. Sub-pixel offset, rotation, scale change invariance, while the contrast and gray-scale changes insensitive, with strong anti-noise.In the process of feature detection and matching using a limited search area, And other optimization strategies to improve the performance of the algorithm.Experiments show that the algorithm has good accuracy and robustness to geometric attacks.The algorithm has been applied to CBERS-0 2B star level 3 data batch automation production, with good application effect.