基于多源传感器数据融合的三维场景重建-《郑州大学学报(工学版)》

文章信息/Info

Author(s):: Chen Yifei; Guo Sheng; Pan Wen’an; Lu Yanhui;; School of Information Engineering, Zhengzhou University; Chinese University of Hong Kong (Shenzhen) Institute of Technology; Shenzhen Institute of Data;

Keywords:: Data fusion; three -dimensional models; deep learning; target detection; feature matching; scene reconstruction

摘要:: LeGO-LOAM算法，将不同类型的特征点进行特征提取与匹配，融合不同时刻的点云完成点云地图的重现。针对构建的点云地图中可能存在的无关目标，借助多源传感器数据和深度学习技术，在三维空间中进行目标检测与剔除。对于点云建模与目标检测两个不同过程，本文采用点云配准的方法对其进行算法融合，最终完成校园环境下的场景重现。本文所提出方法可应用于智慧城市、无人驾驶等领域，具有实际应用价值。

Abstract:: With the development of science and technology, 3D scene reconstruction has important practical value and application prospects in many fields. The use of lidar data to achieve three-dimensional modeling is a widely used technique in recent years, but there are situations such as target redundancy in the reconstruction results in specific scenes. Therefore, the method of camera and lidar data fusion used in this article has nothing to do with specific scenes The target is culled and the 3D scene is reproduced. This paper first uses the lightweight LeGO-LOAM algorithm to extract and match different types of feature points and fuses point clouds at different times to complete the point cloud map reconstruction. For the irrelevant targets that may exist in the constructed point cloud map, with the help of multi-source sensor data and deep learning technology, target detection, and elimination are performed in three-dimensional space. For the two different processes of point cloud modeling and target detection, this article uses the point cloud registration method to algorithmically fuse them, and finally complete the scene reproduction in the campus environment. The method proposed in this paper can be applied to the fields of smart cities and unmanned driving, and it is available for practical application