改进的基于ORB特征的视频实时拼接技术  

Improved Real-time Video Stitching Technique Based on ORB Feature

作　　者：黄剑锋[1] 

机构地区：[1]福州大学机械工程及自动化学院,福建福州350108

出　　处：《机械制造与自动化》2017年第3期212-215,共4页Machine Building & Automation

摘　　要：为了保证在拼接效果的基础上提高视频拼接技术的实时性,选择基于ORB算法实现特征检测,并在此基础上采用了一系列的策略。首先引入了ROI,将待检测的图像序列划分为感兴趣区域和非感兴趣区域;然后采用ORB特征算法和使用二进制串的汉明距离进行图像特征的检测、描述与匹配;再通过参数估计算法PROSAC对特征匹配点对提纯并鲁棒地计算出变换矩阵H,最后采用位置加权平滑算法进行图像融合。实验结果表明,使用的算法速度快,鲁棒性能好,适用于视频实时拼接。To improve the instantaneity of video stitching technology, based on ensuring the effect of stitching, this article chooses the ORB algorithm to implement feature detection, and adopts a series of strategies. Firstly, ROI is used to divide the image se-quence into two parts： interested region and uninterested region. Then ORB algorithm and hamming distance （ in which binary string is used） are used to detect the chart feature and matching feature. Then PROSAC algorithm is used to purify the feature matching point pairs and to calculate the transformation matrix H. Finally, the position weighted smoothing algorithm is adopted in the image fu-sion. The experimental results show that if the algorithm above is used, the robustness is better and the speed is faster, so that it is suitable for real-time video stitching.

关 键 词：视频拼接 ROI ORB算法 PROSAC算法 位置加权平滑算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]