顾及粗差的低空数字影像拼接

低空遥感影像相对于传统航空和卫星影像覆盖面积小且像幅数多,影像拼接是内业处理的重要工作之一。低空摄影机畸变大及飞行平台不稳定等因素易产生同名点匹配定位粗差,可能致使拼接低空影像接边误差较大。为减少这些粗差,利用具有尺度、旋转和平移不变性的SIFT算法进行低空影像匹配,然后采用Huber算法约束粗差,并通过Levenberg-Marquardt非线性最小二乘法进行平差,以获得精确的影像拼接单应矩阵。实验验证了该方法可减少拼接缝,提高了拼接精度。

Low-Altitude Aerial Images Stitching Considering Gross Error

Image stitching is one of the important tasks when processing low-altitude aerial images covering smaller land areas haveing a larger number of images,relative to traditional aerial and satellite data.The gross error for corresponding points caused by camera distortion and instability of the flight platform,may result in a large stitching error on images edges.To eliminate the gross error,the SIFT algorithm with scale,rotation,and translation,was used for images matching.and The gross error was restricted by the Huber algorithm,and the Levenberg-Marquardt nonlinear least square method was used for data adjustmentNext accurate homography matrices where calculated for image stitching.Testing showed that the algorithm can reduce seams and improve image stitching accuracy.