基于CUDA与FAST特征的实时机场全景图像拼接并行算法研究

Real-time Airport Panorama generation based on CUDA and FAST Features

文章提出了一种基于CUDA与FAST特征的全景图实时生成方法。针对相邻图像的无缝拼接,设计了一种基于动态规划缝合线与三角函数的方法。以贵阳机场全景视频监控系统开发过程中采用的全景拼接技术开发过程中的实验,采用了基于四路摄像头图形的全景拼接构成,由于民航业对图像实时性,准确性要求高,实验通过包括柱面投影,图像特征码抽取和图像耦合,图形梯度变化等进行了同步优化整合,实验成果指出通过本实验构建的算法及平台,有效的在保证图像全景生成实时性的基础上,提升了图象耦合部位的可视性和可用性。

This paper proposed a airport panorama generation method based on Fast features and CUDA. For the seamless Mosaic of adjacent images, a method based on dynamic programming suture and trigonometric function is designed. Experiments in this article for the guiyang airport panoramic video monitoring system used in the process of development of panoramic stitching technology development in the process of experiment, using the four road camera based on graphic panoramic stitching, since the civil aviation image real-time and accuracy requirement is high, the experiment by including cylindrical projection, image signature extraction and coupling, optimizing gradient change of synchronization to the graphics design, experimental results point out that through the building algorithm and platform of the experiment, effective in ensuring real-time panoramic image generation is, on the basis of improved image coupling parts of a visibility and availability.