The video mosaicing under no platform information was reseached, and an approach to airborne video mosaicing based on similarity accumulation and bundle adjustment was proposed. The approach uses the basic idea of string matching and a kind of similarity accumulation-based frame matching method to realiz the optimal matching among frames without any information about the platform. Furthermore, in order to eliminate accumulated errors in sequential mosaicing, it uses the bundle adjustment method to estimate the parameters of 2D planar transformations, and a corresponding optimization algorithm can also be given. The performance of this video mosaicing method was testified on some real data and compared with the image composite editor (ICE), a typical general-purpose image mosaicing software. The results show that, as far as airborne video data is concerned, this method is as good as, or even better than ICE.%研究了无平台信息的视频图像拼接,提出了基于相似性积累与光束法平差的机载视频图像拼接方法.该方法借鉴字符串匹配的基本思想,用基于相似性积累的帧匹配方法解决无平台信息下的帧间优化匹配问题;采用基于光束法平差的空间变换参数估计方法消除序贯拼接所可能产生的累积误差,并消除序贯拼接所可能产生的累积误差,并给出了相应的优化算法.利用若干真实机载视频数据对该拼接方法的有效性进行了验证,比较分析了该方法与典型的通用图像拼接软件ICE的实际拼接效果.实验结果表明,对于机载视频数据而言,利用该方法能够获得与ICE相当甚至更优的结果.
展开▼
