基于边缘特征的虚实图像精确配准

3D模型的虚拟图像与可见光图像（实图）的配准可以增强信息的互补性，在通过INS装置粗配准后，其配准问题变成了经典的多模态图像精确配准问题。本文提出了一种基于边缘信息全局最优匹配的3D模型图像与可见光图像精确配准算法。该算法首先基于相位一致性模型提取3D模型图像与可见光图像视觉相似的边缘结构；再采用模拟退火算法搜索求解在边缘配准度最大对应的空间变换模型的最优参数值。4组实验结果表明,本文算法能够充分利用图像间的相近似的边缘结构和模拟退火算法得到配准变换参数,有效地实现多模态图像（不同光照及对比图像对，不同传感器的航拍图片对，虚拟3维视图与可见光图像）的精确配准。

The Precise Registration of Virtual and Real Image by Edge Feature

The registration of virtual and real image can enhance the complementary advantages. After initial registration by INS instruments information, this registration problem is transformed to a classical multi-modal image registration problem. A precise registration algorithm for virtual and real image by edge feature and global optimization was proposed in this article. At first, the edge structure of 3D model and visual image similar with human vision was extracted by the edge algorithm base on phase congruency model. Then the global optimal parameter with respect to maximal registration value of space transforming model was calculated by simulated annealing algorithm. 4 groups of experiments result show that the proposed method can make full use of the similar edge structure between 2 images; simulated annealing algorithm can achieve the better parameter for registration transform. In these experiments the multi-modal images including image pair with different illumination, the air image pair by different sensor, virtual 3D model projection image and visual photo were registered with high accuracy.