遥感影像拼接技术作为影像成图的重要手段，一直是遥感领域研究的热点。在保证拼接精度的同时，如何提高拼接效率，成为低空遥感技术应急应用的难点。针对低空航空遥感影像成像姿态变化剧烈、数据量大等特点，提出针对无序影像改进的基于稀疏矩阵的光束平差SURF拼接算法，即对无序影像进行特征选取和匹配，自行判断相邻和相间位置信息，配准平差中引入稀疏矩阵进行加速。方法在保证精度的同时，极大提高了航空相机影像拼接算法的速度，能很好地解决无人机影像高建筑物的拼接错位问题。无人机倾斜数据集拼接和航空相机影像拼接的结果验证了利用本算法进行高空间分辨率影像拼接的精度和效率。

Remote sensing image mosaic technology, an important means of image mapping, has been the focus of remote sensing research over years. In order to ensure the stitching precision, finding proper methods to accelerate mosaicking efficiency is one of the great challenges in the emergence application with low altitude remote sensing. For the low-altitude aerial remote sensing imagery, the attitude changes drastically, the data volume is large, and so on. An improved SURF splicing algorithm is proposed based on sparse matrix for the disordered image, which selects and matches the disordered image and judges its neighbors by itself. With the interposition position information, the sparse matrix is used to accelerate the registration adjustment. This method can greatly improve the speed of the aerial camera image mosaic algorithm while ensuring the accuracy, and can well solve the problem of the mosaic misplacement of the high-rise buildings of the drone image. The proposed algorithm is then verified by UAV tilt photography image dataset and aerial image dataset, and the mosaicking results demonstrate the merits of our method.