基于改进ORB特征检测的全景视频拼接

PANORAMIC VIDEO STITCHING METHOD BASED ON IMPROVED ORB FEATURE DETECTION

针对全景视频拼接中算法复杂度高且拼接后的视频存在拼接裂缝和"GHOST"现象的问题,提出一种基于改进ORB特征检测的全景视频拼接方法。首先采用多尺度空间来检测特征点以及对检测参数进行设置,使得ORB算子具有尺度不变性且分布均匀;然后采用Hamming距离进行特征匹配并采用RANSAC算法去除误匹配点;最后通过采用复杂度较低的动态规划算法找到最佳缝合线,并对拼接后的图像采用泊松融合进行平滑处理。将Ladybug全景摄像机拍摄的1帧兰州西科站全景视频图像进行仿真实验。仿真结果表明,该全景视频拼接方法在实时性方面表现优异,对存在的拼接裂缝和"GHOST"现象有很好的抑制作用。

Aiming at the high complexity of the panoramic video stitching algorithm and the problem of stitching cracks and 'GHOST'phenomenon,a panoramic video stitching method based on improved ORB feature detection is proposed. Firstly,the multi-scale space is used to detect the feature points and to set the detection parameters,so that the ORB operators are scale invariant and evenly distributed. Then Hamming distance is used for feature matching and RANSAC algorithm is used to remove the mismatch point. Finally,the optimal stitching crack is found by using the low complexity dynamic programming algorithm,and the stitching image is smoothed by Poisson fusion. We take 1 frame panoramic video image of Lanzhou West Railway Station, shot by Ladybug panoramic camera, for simulation experiments. The simulation results show that the panoramic video stitching method is excellent in real-time performance,and has good inhibitory effect on the stitching cracks and 'GHOST'phenomenon.