室内场景的三维感知与理解

项目名称： 室内场景的三维感知与理解

项目编号： No.61272327

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 南亮亮

作者单位： 中国科学院深圳先进技术研究院

项目金额： 82万元

中文摘要： 对现实场景进行感知与理解是计算机图形学和计算机视觉领域研究的重要课题。室内场景的复杂性和多样性决定了场景的数据获取和快速建模都具有极大的挑战性，是图形学领域国际公认的难题。本项目基于大量室内场景的激光扫描点云数据，分析场景中主要物体的几何特征、结构特征，以及多个物体间的相互关联，利用物体的几何和功能的先验知识，实现对场景的识别与理解。在此基础上，利用几何模型库对场景进行快速有效的表达。在对室内场景进行分析时，我们强调物体之间的关联以及物体与环境的上下文关系，以此配合几何结构和先验知识获得物体的功能性描述。本项目旨在探索室内场景的感知、理解与高效表达技术，项目研究将突破传统的基于影响物体识别方法的限制，对计算机图形学从"几何表达"向"语义理解"方向发展产生积极的推动作用。另外，在应用方面，能够为室内场景的数字化奠定理论和技术基础，将极大地推动城市精细化管理和相关服务向智能化、智慧化方向发展。

中文关键词： 室内场景；感知；理解；三维重建；

英文摘要： Recognition and understanding of real world scenes is an important research topic in computer graphics and computer vision. Data acquisition and modeling of indoor scenes is a great challenge due to the complexity and diversity of the indoor scenes. This make it an internationally recognized problem in graphics field. Based on the laser scanned data of indoor scenes, this project aims to develop an array of new methodologies and algorithms for recognizing and understanding complex indoor scenes. We focus on analysis of the geometrical and structural characteristics of indoor objects, as well as intra-relation between multiple objects interrelated in the scenes. Knowledge and constraints are then built on geometry and function of the objects, which are further used to achieve the final scene recognition, reconstruction, and understanding. We emphasize the relationship between objects, which complement the prior knowledge of geometry and functional description. The project strives to bring together methods from computer graphics and computer vision and advances the development of both fields. The project will overcome the limitations of the traditional object recognition methods and steer graphics research from 'geometric representation' to 'semantic understanding'. In addition, the application will greatly promot

英文关键词： indoor scene；recognition；understanding；3D reconstruction；