基于本质矩阵的无人机影像精确匹配

无人机影像以其低成本、获取容易、信息量大等优点得到了广泛的应用。影像匹配是影像数据处理的重要环节，常用影像匹配的方法存在误匹配多或剔除大量正确匹配的问题。LMedS算法比其他稳健方法更严格，"5点算法"得到的本质矩阵可用于匹配点对共面的场景。本文使用LMedS算法结合"5点算法"求解本质矩阵作为剔除误差的模型，同时利用ORB算法提取速度快的特点，构建了一种精准同时兼顾了效率的匹配方法。本文对该算法开展了试验研究，并与其他常规算法进行比较。试验结果表明，本文算法保留的匹配点数数量多，分布较均匀，适用于多种场景，是一种有效、快速精准的影像匹配算法。

An accurate UAV image matching based on essential matrix

Unmanned aerial vehicle (UAV) images have been widely used for their low cost, easy access, and large amount of information. Image matching is an important research topic because the commonly used image matching methods often mismatch or eliminate a large number of matching pair. LMeds algorithm is more stricter than other robust method. The essential matrix obtained by "5-point algorithm" can be used to match point-pair coplanar scenarios. This paper uses the essential matrix, which is solved using the LMedS algorithm combined with the "5-point algorithm", as the model to culling the error. At the same time, the ORB method with the attributes of fast featuring characteristics is used to construct a precise and efficient matching method. The experimental result shows that the proposed algorithm retains a large number of uniformly distributed matching pairs. This method is fast and accurate with efficiency and can be applied in various scenarios.