基于离线标定的全景视频快速拼接算法设计

Design of panoramic video fast stitching algorithm based on offline calibration

针对实际条码识别系统中全场景监控和传统图像拼接算法速度慢的问题,提出了一种基于离线标定的快速全景视频拼接算法。在实际应用中多台相机位置固定,采用离线标定计算出图像拼接的单应性矩阵,在实时拼接中直接加载该矩阵进行计算,从而省去了大量的特征提取和配准时间。为了提高图像特征的配准精度,设计了一种改进的SIFT(Scale-Invariant Feature Transform)特征匹配算法。实验结果表明,该算法对比传统算法显著减少了错误特征匹配的影响,提高了图像拼接速度,具有更好的特征匹配效果。

A fast video stitching algorithm is proposed based on the problem of slow speed of panoramic monitoring and traditional image stitching algorithm in actual barcode recognition system.In practical applications,multiple cameras are fixed in position,and the homography matrix of image stitching is calculated by offline calibration.The matrix is directly loaded in real-time stitching for calculation,thereby a lot of time feature extraction and feature matching time are eliminated.In order to improve feature matching accuracy,an improved SIFT(Scale Invariant Feature Transform)feature matching algorithm is designed.The results show that the proposed algorithm can significantly reduce the influence of error feature matching,improve the image stitching speed and have better feature matching effect.