A Novel Adjustment Model for Mosaicking Low-Overlap Sweeping Images

This paper proposes a novel adjustment model for mosaicking low-overlap sweeping images captured by medium-altitude unmanned aerial vehicle (UAV) with a long focal length. Commonly used methods are not suitable for this type of data. Three innovations are proposed to make this possible: first, building a new error model denoted by an error homograph matrix, which includes the camera parameter error, perspective center error, image attitude error, and the error of projection plane; second, the error homograph matrix is considered an unknown variable and is optimized, instead of optimizing the exterior orientation element in commonly used photogrammetry method; and third, the proposed algorithm is a global optimization, avoiding the cumulative error that appeared in traditional homographic-based methods. The results reveal that the proposed adjustment model can effectively eliminate the misalignments in the seam lines, compared to the direct homograph transformation. The numerical experiment results also demonstrate that the algorithm has perfect convergence and stability. In addition, this method is also suitable for low-altitude UAV images when the covered area can be regarded as a plane.