多RGB-D传感器集成的在线室内高精度三维测图方法

Taking into account the limited FOVs and less visual features caused by monocular RGB-D device, the project propose using multiple RGB-D cameras for indoor 3D mapping. The main research contents include: firstly, in order to avoid signal interference among RGB-D devices, we propose a semi-automatic extrinsic calibration method based on the multi-RGB-D platform without overlapping in their FOVs. Secondly, based on calibration model, pose updating is achieved by integrating 2D, 3D visual and 3D geometric observations detected from multi-sensors. To enforce global consistency, we employ a combination of techniques to enhance the efficiency and accuracy of loop closure detection and then factor graph optimization is used to mitigate trajectory drift; specifically, the graph edges are weighted by residual errors. In this research, the proposed multi-RGB-D SLAM system contains the full mathematical analysis and technical details for camera calibration, pose tracking and map refinement. This work can not only improve the mapping accuracy and enlarge the mapping range of RGB-D devices, and also has great theoretical and practical meanings on RGB-D indoor mapping.

针对现有单RGB-D传感器室内测图存在的视场角受限、特征过少等关键问题，本项目将单RGB-D扩展为多RGB-D，系统研究面向复杂室内环境多RGB-D联合的在线高精度三维测图理论方法,主要内容有：（1）针对RGB-D传感器间存在的信号干扰，研究无重叠视角的多RGB-D平台校正方法；（2）充分利用RGB-D特征观测信息，构建多视角二三维视觉特征与三维几何特征融合的传感器姿态更新模型；（3）集成特征词袋、运动度量的回环概率模型构建与顾及闭环残差的相机轨迹实时平差方法。旨在形成多RGB-D联合在线测图中“平台校正-姿态更新-回环检测-全局优化”完整路线，显著提高RGB-D实时三维测图应用的效能，为RGB-D室内三维测图理论方法和技术应用发展提供直接支撑。

本研究结合VR/AR，智慧城市建设，室内三维应用等需求，以空间地理信息与计算机视觉技术为核心基础，以实现建筑物全生命周期管理以及提升室内场景应急响应能力为目标，提出多Kinect传感器协同测图下的室内BIM自动化三维重建方法，完成了多Kinect传感器协同测图过程中多传感器校正与数据流同步、相机姿态更新及相机轨迹漂移等关键问题的研究。同时研发了一套集室内三维测图到室内三维重建一体化流程的软硬件系统，系统具备大场景室内三维与全景测图能力，室内BIM模型自动化三维重建能力，以及线上VR可视化和空间漫游的能力，系统在2020年深创赛，广东省高校科技成果转化大赛以及深圳市南山区“创业之星”大赛中获得重要奖项。整体上，本项目的研究成果具有重要的科学意义以及广泛的实用价值。