文献类型：期刊文献

中文题名：基于球面空间匹配的双目鱼眼全景图像生成

英文题名：PANORAMIC IMAGE GENERATION WITH DUAL FISHEYE BASED ON SPHERICAL SPATIAL MATCHING

作者：何林飞[1];朱煜[1];林家骏[1];黄俊健[1];陈旭东[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2020

卷号：37

期号：2

起止页码：166

中文期刊名：计算机应用与软件

外文期刊名：Computer Applications and Software

收录：CSTPCD;;北大核心:【北大核心2017】；

基金：上海市科学技术委员会科研计划项目(17DZ1100808)。

语种：中文

中文关键词：双目鱼眼镜头;全景图像;空间匹配;校正;奇异值分解

外文关键词：Dual fisheye lens;Panoramic image;Spatial matching;Correction;Singular value decomposition

摘要：为了利用双目鱼眼采集系统获得360度全景图像,提出一种空间匹配全景图像生成的算法,分为展开和拼接两部分。在展开部分,先将双目鱼眼图像映射到空间单位球面上,再将拼接完成后的全景球面图像根据经纬展开法映射为全景展开图像。在拼接部分,先提取单位球面图像上的特征点并进行匹配,然后利用匹配特征点,通过奇异值分解直接确定空间旋转矩阵,得到空间匹配关系。该算法计算得到的全景展开图像在拼接部分的峰值信噪比达到25dB以上,结构相似性达到0.9以上,通过空间旋转关系匹配后得到的全景图像具有更精确的结果。该算法在Android平台实现,具有良好的应用价值。
In order to obtain 360 degrees panoramic image by using dual fisheye system,this paper proposes a panoramic image generation algorithm based on spherical spatial matching.The algorithm consists of two main parts,unwrapping and splicing.In the unwrapping part,we mapped dual fisheye images to spatial unit sphere,and the panoramic spherical image after splicing to panoramic image was mapped to the panoramic expansion image by latitude-longitude projection.In the splicing part,the feature points on the unit spherical image were extracted and matched.Then,the space rotation matrix was directly determined by singular value decomposition,and the space matching relationship was obtained.The peak signal-to-noise ratio of the panoramic expansion image calculated by this algorithm was more than 25 dB,and the structural similarity was more than 0.9.The panoramic image obtained by spatial rotation matching has more accurate results.The proposed method is implemented on the Android platform,and it has good application value.