改进SIFT算法的小型无人机航拍图像自动配准

Unmanned Aerial Vehicle Serial Aerial Image Automatic Registration Based on Improved SIFT Algorithm

针对小型无人机航拍图像视点离散、视角变化有一定运动规律的特点,首先对航拍图像进行数据预处理,结合Harris特征点和SIFT特征向量的优势,提取Harris特征点、计算特征点的特征半径和SIFT特征向量,并利用PCA降低特征向量的维数;然后采用最邻近(NN)方法进行特征匹配,利用BBF算法搜索特征的最邻近以提高匹配速度;最后采用PROSAC算法提纯特征点匹配对并精确计算运动模型参数,实现了图像的自动配准。实验证明,该图像配准方法在准确性、效率方面较经典的SIFT算法有较大的提高。

Due to the disperse and regular of view points and the view angle of UAV Aerial Image,the image data was preconditioned at first,then the Harris feature points with SIFT feature vectors were combined,Harris feature points were extracted,the characteristics radius of feature points and SIFT feature vector was calculated,and PCA（Principal Component Analysis） was used to reduce the dimension of SIFT feature vectors.And then the most close method（NN） was used to feature matching,the BBF algorithm was applied to search the nearest neighbor feature for improving the matching speed.Finally,the PROSAC algorithm was used to purify initial feature point matching pairs,and motion model parameters were calculated,the image automatic registration was achieved.The results of experiment proved that such algorithm was more efficient and exact than the classic SIFT algorithm.