一种概略位置与姿态条件下的无人机影像拼接方法

A method of UAV image mosaic under the condition of approximate position and posture

为解决现有无人机影像拼接方法对位置与姿态信息需求高、对拍摄条件要求严和不易满足拼接精度要求的问题,本文提出一种新的无人机影像拼接方法。该方法先构造一个特征对应能量项,使匹配特征对之间的平方距离之和最小,以几何方式对齐影像;再建立一个正则化项,通过保持所有变换尽可能刚性来直接约束影像变换参数,以避免最终拼接中的全局失真。试验结果表明,本文提出的影像拼接方法能有效消除直接、间接定向过程所产生的误差,处理速度显著加快,拼接精度是基于Capel影像拼接方法的二倍。

The existing UAV image mosaic methods require high demand for position and posture information,strict requirements for shooting conditions and difficult to meet the mosaic accuracy.In order to solve the problems above,this paper proposes a new UAV image mosaic method.In the method,a feature corresponding energy term is constructed to minimize the sum of square distances between mosaic feature pairs and align the image geometrically.Then a regularization term is established to directly constrain the image transformation parameters by keeping all transformations as rigid as possible to avoid global distortion in the final mosaic.The experimental results show that the image mosaic method proposed in this paper can effectively eliminate the errors caused by direct and indirect orientation processes,significantly accelerate the processing speed,and the mosaic accuracy is twice that of Capel image mosaic method.