目的 在计算机图形学发展的过程中,数字三维场景长期对于学术界和工业界都起着至关重要的作用。其作用体现在展示渲染效果、支持应用环境、充当交互载体等多个方面。然而,三维场景本身作为一种数据形式,其结构复杂且没有统一的数据结构,故难以像诸如图像、文本数据集一样被大量获取与应用。所以,已有一些工作尝试让计算机自动生成场景或是让计算机辅助生成场景。而在众多场景之中,室内场景尤为重要。本综述总结归纳了这些数字室内三维场景生成工作。方法 本综述提出从三个方面调研和总结场景构建的主要工作：场景自动构建、基于用户交互的场景辅助构建、基于多通道与丰富输入的场景构建。自动构建在于让计算机直接基于当前三维内容给出场景构建结果；交互构建在于让用户控制计算机并辅助生成场景；多通道构建在于让生成的场景参考输入的图像、文本、点云。最后,本文总结了科研工作的应用场景和关键技术,并介绍了一些其他应用情景与未来面对的诸多挑战。结果 我们利用本篇综述对三维场景构建的研究和应用进行了总结归纳。结论 数字室内三维场景构建的发展前景十分广阔。在未来,随着新算法的不断提出以及三维场景数据集日益完善,数字室内三维场景构建领域也将持续发展。

In the process of developing computer graphics, digital 3D scenes have played a vital role in academia and industry. It can show the rendering result, be the environment of application, and provide a foundation for interacting. However, as a data structure, a digital 3D scene has a complex structure, and no unified consensus has been given subject to its data structure. Thus, digital 3D scenes are difficult to be obtained and applied in large quantities compared with image and text datasets. Consequently, some existing work tries to let computer automatically generate 3D scenes or interactively help generate scenes. This survey investigates and summarizes 3D digital scene generation methodologies in three aspects: automatic scene generation, aided scene generation based on user interaction, and scene generation with multi-channel and rich input. The automatic generation is to let the computer directly build results based on the 3D content. The interactive generation is to let users control the process of computer-aided scene generation. The generation with multi-channel and rich input is to make the generated scene refer to the input image, text, point cloud, etc. Eventually, this survey will summarize the application scenario and core technology of the papers and introduce several other typical application scenarios and future challenges.