一种快速的基于SIFT算法的无人机航拍序列图像自动拼接方法

无人驾驶飞机在短时间内具有相对稳定的航向、飞行速度、飞行高度等特点，这些特点决定了相同时间间隔的航拍序列图像的重合区域相对稳定，在此基础上，本文提出了航拍序列图像拼接方法的约束性条件，及一种快速的无人机航拍序列图像的自动拼接方法。该方法运用自回归模型，将历史相交区域中心点作为模型的输入数据，对当前用于拼接的重合区域进行预测，然后在预测区域内进行SIFT特征提取和特征提纯，并根据提纯后的特征进行快速的序列图像拼接。通过在无人驾驶飞机航拍序列图像上对该方法性能进行实验，结果显示本文的方法具有较好的拼接效果，在海量无人机航拍数据的情况下，能大大减少拼接时间。 

A Fast Automatic Mosaic Method on Unmanned Aerial Vehicle Images Using the SIFT Algorithm

Unmanned aerial vehicle has many characteristics, such as the relatively stable aerial navigation, flight speed, flight height. These characteristics decide that there are relatively fixed positions between the neighbor images. Based on these characteristics of the unmanned aerial vehicle (UAV) images, the constraint condition of the mosaic method on the UAV images is proposed, and then a fast Automatic Mosaic Method of the UAV images is proposed. Firstly, the proposed method use the AR model which use the center position of the history overlap areas as the input data to forecast the overlap areas of the two images where the SIFT features are extracted, and then mosaic the two image using the extracted SIFT features. The results of experiments on the UAV images show that the method has good performance on the mosaic image, and the time cost reduced largely.