全景视频摄像机的非参数化几何校正与拼接

提出了一种新的多相机全景视频拼接几何校正拼接方法。先用全景视频摄像机组采集标准标定场的稀疏特征点阵,利用自适应网格细分技术把采集的图像和标定场坐标细分到像素级对应,则全景视频的几何校正和拼接可以通过直接映射得到。在摄像机阵列相对位置固定时,可以一次校正多次使用。本文方法避免了普通拼接方法的特征点检测、匹配优化等参数求解过程,尤其是采用具有较大畸变的广角镜头和相邻相机视场重叠较少时更能体现出其优越性,并可以很容易地迁移到类似的应用中。通过一个原型系统,验证了本文方法的有效性。

Non-parametric geometric correction and mosaic of panorama camera

This paper presents a novel geometric correction method for panorama camera.Firstly,the sparse feature point grid of a standard calibration field is captured by the grouped cameras.And then the correspondence between image pixels and panorama coordinates is found by using adaptive mesh subdivision,which naturally produces a panorama video.Most of current methods are based on the assumption that the relative positions of the cameras in the group are fixed.The method of this paper doesn′t need the common routine of feature detection,key points registration and parameter optimization.It fits for the case that the lenses have great distortion and lenses overlap regions are small or not.Once the camera is corrected,it doesn′t need to repeat the process until the cameras are moved relatively.It can be easily reused in the similar situation where the lenses have nonlinear distortion,such as microscope.This method is confirmed experimentally with a prototype system in different conditions.