Aiming at the problem of error accumulation in aerial array remote sensing image stitching algorithms, a method for eliminating cumulative errors in aerial array image stitching was proposed. First of all, the attitude information and position information of the aerial remote sensing platform were used to complete the projection transformation of each image and rough image stitching. The independence of image’s error in stitching after each projection transformation was ensured. Thereby the error accumulation in image stitching was avoided. Then, the SIFT algorithm was used to extract the image matching pairs point in the overlapping area of the rough stitching adjacent images. According to the position difference between matching pairs of adjacent images, the principle of error sharing was adopted, and independent slightly adjust and coordinate fine-tuning were performed one by one. Until a certain accuracy was met, the accuracy of image stitching was improved and the accumulation of errors was avoided at the same time. Finally, the image quality was evaluated by combining with subjective evaluation (Compare the geographic correspondence between the corrected satellite image and the stitched image in this article) and the objective evaluation (in SSIM algorithm). The theoretical analysis and experimental results show that this method can effectively avoid the problem of error accumulation in aerial array remote sensing image stitching. Compared with the current main splicing methods, this method can maintain consistency with the two-dimensional coordinates of geographic space, and has good practical value and application prospects.