The construction of digital maps typically relies on manual stereoscopic plotting based on aerial imagery, which demands considerable time and cost. As a result, it is challenging to promptly reflect frequent building changes associated with urban development in digital maps. In this study, to automate the modification and updating of digital maps, we proposed an automated segment anything model (SAM) 2-based building segmentation approach that utilizes You Only Look Once (YOLO) v8 for building detection and image processing techniques to extract building boundaries from ortho-images. In the proposed methodology, we were able to automatically generate prompts for SAM2 by applying image processing techniques to the bounding boxes of buildings detected by YOLOv8, removing noise and creating clear masks. Furthermore, through performance comparison experiments between the pretrained and fine-tuned SAM2, we found that the fine-tuned SAM2 significantly improved building segmentation performance because of the additional training specialized for building data. In experiments on comparing a single prompt and multiprompt inputs, we observed that multiprompt inputs enabled a more precise and accurate building segmentation, confirming that prompts play a crucial role in enhancing model performance.