GeoGaussian is a geometry-aware Gaussian Splatting method that enhances scene rendering by preserving accurate geometry. The method introduces a novel pipeline to initialize thin Gaussians aligned with surfaces, using a carefully designed densification strategy to maintain scene geometry and texture. By leveraging explicit geometric information, GeoGaussian ensures that Gaussians lie on smooth surfaces, improving rendering quality in structured regions. The method is evaluated on public datasets and achieves state-of-the-art performance in novel view synthesis and geometric reconstruction. Key contributions include a parameterization with explicit geometry meaning for thin 3D Gaussians, a geometrically consistent constraint to align thin Gaussians with smooth surfaces, and improved rendering quality compared to existing Gaussian Splatting methods. The method is compared with 3DGS and LightGS, showing superior performance in rendering quality, especially in low-textured regions. GeoGaussian also demonstrates robustness in sparse view scenarios and outperforms other methods in structured environments. The method uses geometry-aware constraints during initialization, densification, and optimization to enhance scene geometry and rendering performance. However, the method has limitations, such as performance degradation in scenarios with few smooth regions and potential issues with incorrect geometric residuals. Overall, GeoGaussian provides a more accurate and efficient approach to Gaussian Splatting for scene rendering.GeoGaussian is a geometry-aware Gaussian Splatting method that enhances scene rendering by preserving accurate geometry. The method introduces a novel pipeline to initialize thin Gaussians aligned with surfaces, using a carefully designed densification strategy to maintain scene geometry and texture. By leveraging explicit geometric information, GeoGaussian ensures that Gaussians lie on smooth surfaces, improving rendering quality in structured regions. The method is evaluated on public datasets and achieves state-of-the-art performance in novel view synthesis and geometric reconstruction. Key contributions include a parameterization with explicit geometry meaning for thin 3D Gaussians, a geometrically consistent constraint to align thin Gaussians with smooth surfaces, and improved rendering quality compared to existing Gaussian Splatting methods. The method is compared with 3DGS and LightGS, showing superior performance in rendering quality, especially in low-textured regions. GeoGaussian also demonstrates robustness in sparse view scenarios and outperforms other methods in structured environments. The method uses geometry-aware constraints during initialization, densification, and optimization to enhance scene geometry and rendering performance. However, the method has limitations, such as performance degradation in scenarios with few smooth regions and potential issues with incorrect geometric residuals. Overall, GeoGaussian provides a more accurate and efficient approach to Gaussian Splatting for scene rendering.