3D Gaussian Splatting (3DGS) has emerged as a prominent technique for 3D representations, offering efficient training and real-time rendering of novel views. This survey aims to provide a comprehensive overview of 3DGS, including its tasks, technologies, challenges, and opportunities. The primary objectives are to help newcomers understand the field and assist researchers in organizing existing technologies and challenges. The survey categorizes 3DGS into optimization, application, and extension, and identifies nine types of technical modules. It also examines common challenges and proposes potential research opportunities. The introduction highlights the advantages of 3DGS over Neural Radiance Fields (NeRF) in terms of efficiency and controllable explicit representation. 3DGS is applied in various fields, including VR/AR, robotics, film, and urban planning. The survey is structured into sections covering background, optimization, applications, extensions, and future research directions. Optimization sections focus on efficiency, photorealism, and generalization. Techniques include vector quantization, pruning, and hardware acceleration. Photorealism improvements involve joint 2D-3D training, frequency-domain guidance, and reflective surface reconstruction. Generalization efforts address sparse view settings and multi-view stereo. Applications sections detail 3DGS's use in human reconstruction, AIGC, and autonomous driving. Human reconstruction includes body and head reconstruction, with methods like SMPL and FLAME. AIGC applications include text-to-3D object generation, image-to-3D object generation, multi-object and scene generation, and 4D generation. Autonomous driving applications focus on dynamic scene reconstruction and SLAM. The survey concludes by emphasizing the limitations of existing research and proposing avenues for future advancements. An open-source project is also mentioned, providing a resource for researchers to access the latest research information.3D Gaussian Splatting (3DGS) has emerged as a prominent technique for 3D representations, offering efficient training and real-time rendering of novel views. This survey aims to provide a comprehensive overview of 3DGS, including its tasks, technologies, challenges, and opportunities. The primary objectives are to help newcomers understand the field and assist researchers in organizing existing technologies and challenges. The survey categorizes 3DGS into optimization, application, and extension, and identifies nine types of technical modules. It also examines common challenges and proposes potential research opportunities. The introduction highlights the advantages of 3DGS over Neural Radiance Fields (NeRF) in terms of efficiency and controllable explicit representation. 3DGS is applied in various fields, including VR/AR, robotics, film, and urban planning. The survey is structured into sections covering background, optimization, applications, extensions, and future research directions. Optimization sections focus on efficiency, photorealism, and generalization. Techniques include vector quantization, pruning, and hardware acceleration. Photorealism improvements involve joint 2D-3D training, frequency-domain guidance, and reflective surface reconstruction. Generalization efforts address sparse view settings and multi-view stereo. Applications sections detail 3DGS's use in human reconstruction, AIGC, and autonomous driving. Human reconstruction includes body and head reconstruction, with methods like SMPL and FLAME. AIGC applications include text-to-3D object generation, image-to-3D object generation, multi-object and scene generation, and 4D generation. Autonomous driving applications focus on dynamic scene reconstruction and SLAM. The survey concludes by emphasizing the limitations of existing research and proposing avenues for future advancements. An open-source project is also mentioned, providing a resource for researchers to access the latest research information.