一种基于自适应三角剖分的航空多光谱影像配准方法

An Aerial Multi - spectral Image Registration Method Based on Adaptive Triangulation

本文针对DMZⅡ相机多光谱影像视场大、多通道影像间形变差异较大的特点,提出了一种高精度影像配准方法。基本流程是:首先,采用Harris算法提取影像特征点,并采用由粗到精和分块匹配的策略获取初始同名点对;然后,在三角网条件约束下,逐步实现整个重叠视场范围的配准;最后,基于小面元微分纠正实现了航空多光谱影像间的髙精度配准。试验表明,该方法能够实现航空多光谱影像共同视场全覆盖,配准误差均小于1个像素,配准精度达到了亚像素级,满足多光谱影像间高精度配准的处理要求。

Aiming at the characteristics of large field view and great deformation differences among multi - channel images of DMZ Ⅱ camera, a high precision image registration method is proposed in this paper. The basic process is as follows: firstly, the image feature points are extracted by Harris algorithm, and the initial tie point pairs are obtained by the strategy which is from coarse to precious and block matching. Then the registration of the entire overlapping field view is realized step by step with the constraint of triangular network. Finally, the high accuracy registration among aerial multi - spectral images is realized based on facet differential correction. Experiments show that the proposed method can achieve full coverage of common field view of aerial multi-spectral images, and the registration errors are less than 1 pixel, and the accuracy reaches sub - pixel level, which meets the processing requirements of high-precision registration among multi - spectral images.