UGVs face significant limitations in environmental perception due to workspace constraints. Integrating UAVs’ advantages in perspective to achieve aerial-ground collaboration has become a mainstream trend. This paper proposes a Gaussian Splatting SLAM system for aerial-ground collaboration. Given that the aerial perspective of a drone and the ground-level perspective of a ground robot are orthogonal, the core idea of this method is to address the coordinate system transformation problem between such aerial-ground orthogonal viewpoints, enabling map fusion through a coordinate transformation matrix. In the visual-inertial front-end of this system, RGB images are processed through an optical flow prompt generator, segmentation module, and uncertainty estimation processing module to extract scene geometry and pose information. Based on this output, the mapping module incrementally constructs and maintains a 2D Gaussian map. To ensure global consistency in large-scale scenarios, the loop closure detection module leverages the novel view synthesis (NVS) capability of Gaussian Splatting for loop detection and Gaussian map correction. Additionally, to address inevitable dynamic object interference in real outdoor scenes, a real-time instance segmentation algorithm (YOLACT++) is employed for processing. The system demonstrates superior performance in map construction and rendering quality compared to most existing methods. In summary, this system enables outdoor operation and supports monocular Gaussian SLAM for large-scale environments up to kilometer-level scales.