一种新的无人机航拍序列图像快速拼接方法  

A New Fast Automatic Mosaic Method on Unmanned Aerial Vehicle Images

作　　者：李岩山[1] 裴继红[1] 谢维信[1] 李良群[1] 

机构地区：[1]深圳大学ATR国防科技重点实验室,广东深圳518060

出　　处：《电子学报》2012年第5期935-940,共6页Acta Electronica Sinica

基　　金：广东省科技推广计划项目(No.A2009007-019(a));深圳市重点实验室提升计划项目(No.CXB200903090020A);深圳市科技计划项目(No.JC201005250040A)

摘　　要：无人驾驶飞机航拍序列图像的邻近帧间含有大量的相交区域,这些相交区域是进行基于特征的图像拼接的基础,但是也成为了降低拼接效率的冗余信息.因此,本文提出一种既稳定又具有较小时间开销的无人机航拍序列图像的自动拼接方法.利用图像拼接过程中查找到的匹配特征点,自适应提取出适于航拍序列图像快速拼接的关键帧;在分析了序列图像的拼接时的相交区域的运动模型的基础上,建立了无人机航拍序列图像配准过程中的特征搜索区域预测的卡尔曼滤波器,减少了特征搜索和配准的时间;然后给出了详细无人机航拍序列图像的快速拼接方法,实验结果显示,本文的方法具有较好的拼接效果和拼接效率.The adjacent frames of the unnmnned aerial vehicle images contain much overlap areas. These areas are the basis of the image mosaic, but also decrease the efficient of the mosaic. This paper proposed a fast and stable automatic mosaic method on unmanned aerial vehicle images. A new mosaic-oriented frame selection method by using the feature points which were found in the process of the mosaic to select the most feasible flames for mosaic was proposed; based on analysis of the movement model of the overlap areas in the images,a Kalman filter for predicting the search area of feature points was proposed to reduce the time-cost in the process of searching and matching features. And then, the detail steps of the algorithm were presented. The experimental results showed that the proposed algorithm can finish the seamless image mosaic with low time-cost.

关 键 词：无人机航拍序列图像 自动拼接 卡尔曼滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]