基于控制点网格优化的航拍图像拼接算法

AERIAL IMAGE STITCHING ALGORITHM BASED ON CONTROL POINT GRID OPTIMIZATION

由于航拍图像拼接过程中的特征点误匹配和重影,传统方法使用全局单应性矩阵变换得到的拼接效果并不理想,对此提出一种基于控制点网格优化的图像拼接算法。针对特征点的误匹配,提取输入图像的LIFT特征点,使用RANSAC对特征点进行筛选和匹配,优化特征点对应匹配关系,并给出一种控制点筛选算法;使用控制点结合最小二乘网格约束对图像重叠区域进行网格划分和调整优化,并给出平滑过程中的误差计算方程,从而得到更加精确的图像配准模型;使用加权泊松融合降低色差、淡化拼接缝隙,得到航拍图像的拼接结果。该算法降低了拼接过程中的误配准和重影现象,得到了更精确逼真的航拍图像拼接结果。

Aiming at the mismatching and ghosting of feature points in the process of aerial image stitching,the stitching effect obtained by the traditional method of using global homography matrix transformation is not ideal.We propose an image stitching algorithm based on control point grid optimization.For the mismatching of feature points,we extracted LIFT feature points from the input image,used RANSAC to filter and match the feature points,optimized the matching relationship of feature points,and gave a control point filtering algorithm.Then,the control points combined with the least square grid constraint were used to divide and adjust the overlapping area of the image,and the error calculation equation in the smoothing process was given to obtain a more accurate image registration model.We used weighted poisson fusion to reduce chromatic aberration and lighten the stitching gap,and obtained the stitching result of aerial images.The algorithm reduces the misregistration and ghosting phenomenon in the splicing process,and obtains more accurate and realistic aerial image splicing results.