Firstly,for the acquisition of large-area feature-insignificant ground images,we propose the acquisition of large-area ground images based on multiple cameras because the traditional single camera is affected by the field of view and resolution as well as the external environment.Secondly,we propose to improve the framework structure of unsupervised learning image stitching to improve the image stitching quality.Finally,the stitched images before and after the improvement are evaluated with the help of the evaluation index of image stitching.The experimental results show that the improved method can not only effectively solve the artifacts and distortions generated in the traditional method based on the ground image stitching with insignificant features and high texture similarity,but also solve the changes of the structure generated in the unsupervised image stitching process.The quality of the improved stitching method in this paper is better than that of the traditional method of stitching,and the improved algorithm is highly migratory and can be widely used for stitching a large range of large baseline images in different scenes,not only for a large range of ground without significant features