针对云台网络摄像机监控系统,提出一种基于摄像机视频流的全景图生成算法,以构建更大的监控场景.根据帧间重叠区域的大小选取关键帧,进行柱面投影,利用计算性能优越的SURF(Speeded Up Robust Features,加速鲁棒性特征)算法对所选取的关键帧进行特征点提取,使用基于哈希映射的特征点匹配算法加快特征点的匹配,并结合RANSAC(RANdom SAmple Consensus,随机抽样一致)算法剔除误匹配,估计关键帧之间的变换关系.实验结果表明,该方法能较好实现视频序列的快速拼接,鲁棒性强,具有较高的实用价值.%In order to build a network Camera surveillance system, a panorama mosaic algorithm based on video sequence captured by network Camera is proposed. Firstly, the algorithm selects key frames according to the size of overlapping area between relative frames and projects the key frames to cylindrical separately. Then, abstracting features from the key frames uses SURF method with superior computational performance. For the feature matching aspect the algorithm uses a matching method based on hash mapping to accelerate the matching progress, and then a RANSAC algorithm is used to eliminate outliers and estimate the transformation parameters between key frames. Experimental results show that this method with strong robustness achieve fast mosaic of video frame sequence and has highly valuable in practice.
展开▼
