A survey on 3D Gaussian Splatting (GS) explores its emergence as a transformative technique in explicit radiance field and computer graphics. Unlike traditional neural radiance fields (NeRF), which use implicit, coordinate-based models, 3D GS employs millions of learnable 3D Gaussians for explicit scene representation, enabling real-time rendering and unprecedented editability. This paper provides a systematic overview of recent developments and key contributions in 3D GS, highlighting its practical applicability, performance, and challenges. It compares leading 3D GS models across benchmark tasks and identifies future research directions. The survey emphasizes the unique blend of differentiable pipelines and point-based rendering in 3D GS, which allows efficient computation and rendering by leveraging explicit, structured data storage. 3D GS offers real-time rendering capabilities without compromising visual quality, opening possibilities for applications in virtual reality, augmented reality, and real-time cinematic rendering. Its explicit scene representation provides flexibility for controlling objects and scene dynamics, crucial for complex scenarios. The survey also discusses data-efficient, memory-efficient, photorealistic, and improved optimization methods for 3D GS, along with applications in robotics, dynamic scene reconstruction, AI-generated content, autonomous driving, and medical systems. 3D GS has shown promise in various domains, including SLAM, dynamic scene reconstruction, and AIGC, with potential for further advancements in scene understanding and representation. The survey concludes by highlighting the transformative potential of 3D GS in shaping future developments in computer vision and graphics.A survey on 3D Gaussian Splatting (GS) explores its emergence as a transformative technique in explicit radiance field and computer graphics. Unlike traditional neural radiance fields (NeRF), which use implicit, coordinate-based models, 3D GS employs millions of learnable 3D Gaussians for explicit scene representation, enabling real-time rendering and unprecedented editability. This paper provides a systematic overview of recent developments and key contributions in 3D GS, highlighting its practical applicability, performance, and challenges. It compares leading 3D GS models across benchmark tasks and identifies future research directions. The survey emphasizes the unique blend of differentiable pipelines and point-based rendering in 3D GS, which allows efficient computation and rendering by leveraging explicit, structured data storage. 3D GS offers real-time rendering capabilities without compromising visual quality, opening possibilities for applications in virtual reality, augmented reality, and real-time cinematic rendering. Its explicit scene representation provides flexibility for controlling objects and scene dynamics, crucial for complex scenarios. The survey also discusses data-efficient, memory-efficient, photorealistic, and improved optimization methods for 3D GS, along with applications in robotics, dynamic scene reconstruction, AI-generated content, autonomous driving, and medical systems. 3D GS has shown promise in various domains, including SLAM, dynamic scene reconstruction, and AIGC, with potential for further advancements in scene understanding and representation. The survey concludes by highlighting the transformative potential of 3D GS in shaping future developments in computer vision and graphics.