基于三维旋转模型的红外全景图像拼接方法

分布孔径红外系统(DAIRS)是目前各国在红外搜索跟踪系统领域的重点发展方向,DAIRS系统为了减少冗余,通常减少重叠区域面积以获得较宽视场覆盖,这使得基于特征的拼接方法提取特征数量不足。本文通过构建三维旋转模型,采用基于梯度下降的图像配准方法,并在时间维度上进行多帧迭代来进行三维重建,再采用改进的亮度调整算法和多分辨率过渡区融合消除拼接痕迹。实验结果表明,本文算法对于平面投影和柱面投影都实现了全景图像的无缝拼接,过渡区通过多分辨率融合有利于进步消除模糊现象。

Panoramic stitching of infrared image based on three-dimensional rotation model

:Distributed aperture infrared (DAIRS) system is the key development direction of infrared search and tracking systems in various countries. To reduce the redundancy,the DAIRS system usually reduce the overlap area to obtain wide field coverage,which leads to the shortage of the extracted features in the process of image stitching based on feature method. In this paper,by building three-dimensional rotation model and applying image registration based on gradient descent,three-dimensional reconstruction is done with iteration of multi-frame iteration on the dimension of time. Then the improved brightness adjustment algorithm and multi-resolution fusion of transition region are applied to eliminate the stitching traces. Experimental results show that the algorithm achieves seamless stitching of panoramic images for plane projection and cylindrical projection,and multi-resolution fusion of transition areas is beneficial to eliminate the fuzzy phenomenon.