Abstract: Urban aerial images have extensive applications in urban planning, land management, environmental monitoring, and infrastructure development. Addressing the issue that higher drone flight altitudes lead to increased image capture costs and lower image quality or visibility, this study utilizes low-altitude UAVs for extensive image acquisition. SIFT algorithm is proposed to tackle the poor matching stability and alignment quality issues of the classical SIFT image stitching algorithm. By extracting an image pyramid model, the stability of matching is enhanced. The RANSAC algorithm is employed to reduce outliers and improve image alignment quality. Additionally, a hybrid averaging weighted method is used to eliminate seam artifacts in overlapping areas, ultimately achieving rapid and precise image stitching. Simulation results demonstrate that the improved SIFT algorithm performs well in both stitching stability and quality, producing satisfactory and complete stitched images.