乡村振兴背景下海绵城市虚拟全景影像快速重建研究

Research on the rapid reconstruction of sponge city virtual panoramic image under the background of rural revitalization

针对传统三维重建技术存在稀疏点云数据特征提取精度低,导致对海绵城市虚拟全景影像无法快速重建的问题。提出一种基于乡村振兴背景下的海绵城市虚拟全景影像快速重建方法。此方法采用基于无人机影像的分段式恢复结构算法进行影像特征提取和场景稀疏点云重建;之后根据无人机影像分组顺序对各分组重建结果进行拼接合并,以实现三维空间点和相机位姿优化,最终构建更精准的场景点云,实现全景影像快速重建。实验结果表明,相较于传统的openMVG点云重建方法,本方法的点云重建数量为openMVG的4倍,且重建时间比openMVG方法快了2.91倍。采用两种方法对海绵城市真实场景进行全景影像重建后可知,本方法的残差平均值为0.67px,比openMVG方法的残差减小了0.12px,精度提升了15.19%。综合分析可知,本方法具备较高的重建精度,可有效减少误差累积,消除漂移,从而精准进行海绵城市虚拟全景影像快速重建。

In view of the traditional 3 d reconstruction technology,the feature extraction accuracy of sparse point cloud data is low,which leads to the rapid reconstruction of the virtual panoramic image of sponge city.A rapid reconstruction method of virtual panoramic image of sponge city based on the background of rural revitalization is proposed.This method adopts image feature extraction and sparse cloud reconstruction.Later,the reconstruction results of each group are assembled and merged according to the grouping order of the UAV image to realize the optimization of 3-dimensional space points and camera position,and finally build a more accurate cloud to realize the rapid reconstruction of panoramic image.The experimental results show that compared to the traditional openMVG point cloud reconstruction method,the number of this method is 4 times that of openMVG,and the reconstruction time is 2.91 times faster than the openMVG method.After using the two methods to reconstruct the real scene of sponge city,it can be seen that the average residual value of this method is 0.67 px,which is 0.12 px lower than that of openMVG method,and the accuracy is improved by 15.19%.Comprehensive analysis shows that this method has high reconstruction accuracy,which can effectively reduce error accumulation and eliminate drift,so as to accurately carry out rapid reconstruction of virtual panoramic image of sponge city.