基于ORB和改进RANSAC的无人机遥感图像配准算法

UAV remote sensing image registration algorithm based on ORB and improved RANSAC

针对无人机遥感图像配准对于实时性和稳健性的要求,提出了一种基于ORB(Oriented FAST and rotated BRIEF)和改进RANSAC的无人机遥感图像配准算法。首先通过ORB算法快速进行特征检测和特征描述;接着使用正反双向匹配和余弦相似度方法进行特征点粗匹配;最后,结合空间一致性检测理论、PROSAC(Progressive sampling consensus)算法和L-M(Levenberg-Marquardt)算法来综合改进传统RANSAC(Random sample consensus)算法,对匹配点对进行了提纯并计算出变换矩阵。实验结果表明,本文改进算法的匹配正确率和配准精度分别达到了97.8%和0.563,较ORB+RANSAC算法提高了7.5%和20.6%,较ORB+PROSAC算法提高了4.3%和10.1%。在提纯耗时方面,改进RANSAC算法耗时最短,约占传统RANSAC算法耗时的68.3%和PROSAC算法耗时的90.1%,有效提高了无人机遥感图像配准工作的速度和精度。

In order to meet the requirements of UAV remote sensing image registration for real-time and robustness,a UAV remote sensing image registration algorithm is presented based on ORB(Oriented FAST and rotated BRIEF)and improved RANSAC.An ORB algorithm is used to quickly complete feature detection and description.Subsequently,the forward and backward bidirectional matching algorithm combined with cosine similarity method is used to complete rough matching of feature points.Spatial consistency detection theory combined with PROSAC(Progressive sampling consensus)algorithm and L-M(Levenberg-Marquardt)algorithm is introduced to comprehensively improve the traditional RANSAC(Random sample consensus)algorithm so as to achieve the purification of matching point pairs and the calculation of transformation matrix.Experimental results show that the matching accuracy and registration accuracy of the proposed algorithm in this study reach 97.8%and 0.563 respectively,which are 7.5%and 20.6%higher than the ORB+RANSAC algorithm,4.3%and 10.1%higher than the ORB+PROSAC algorithm.In terms of purification time,the improved RANSAC algorithm takes the shortest time,accounting for about 68.3%of the traditional RANSAC algorithm and 90.1%of the PROSAC algorithm.These results indicate that the proposed algorithm can effectively improve the speed and accuracy of UAV remote sensing image registration.