基于特征的航空视频序列拼接方法

针对航空侦察视频序列，提出了一种基于特征点的视频序列拼接方法，论述了该方法的流程和模块。该方法矫正了航空视频序列中普遍存在的径向扭曲；改进了SIFT特征提取和匹配策略，缩短了算法的时间；提出了改进的RANSAC算法用于剔除匹配外点，提高了视频帧图像之间的变换精度；提出双频带融合技术消除边缘跳变和模糊叠影现象，改善了视频帧融合的视觉效果。实验结果表明，该方法能够有效地实现航空视频序列的近实时拼接，同时保持对噪声、光照和目标运动鲁棒。

Feature based mosaic method for aerial video sequences

A feature based mosaic method was proposed for aerial video sequences, and the flow and modules of the method were addressed. The pervasive radial distortion in aerial video sequences is rectified. The methods of extraction and matching of the SIFT features are improved, and the time of the algorithm is saved. The outliers are kicked out by this improved RANSAC, which enhances the precision of the transformation among frames. Finally, the visual effect of mosaic is increased by the proposed double band blending technique, which eliminates the discontinuous edge and blurring. Experiments prove that the method could realize mosaic in nearly real time, and is robust to image noise, illumination variations and moving objects.