影像数量对SFM三维重建模型测量精度的影响分析

在运动恢复结构(Structure From Motion,SFM)技术的实际应用中,不同数量影像数据集的SFM重建精度和效率通常存在差异。为了研究影像数量对SFM重建模型精度的影响,为三维重建应用中的影像数据采集与选取提供参考依据,利用手持数码相机对目标物体进行数据采集,选取不同数量的影像执行SFM点云和网格模型的重建,并将各组SFM三维模型与激光扫描模型进行模型精确度、噪声和完整度的对比分析。实验结果表明,影像数量与SFM三维模型的精度之间具有一定的相关性,影像数量越多,重建模型的完整度越高,但受SFM特征点匹配误差、配准误差以及影像中多余场景的影响,影像数量的增加会导致模型精确度的降低以及噪声程度的加重。在实际的工程应用中,需要深入分析生产三维模型的特点和需求,选取合适影像数量的数据集进行SFM三维重建。

Analysis on the influence of images on measurement accuracy of three-dimensional model reconstructed by SFM

In the practical application of structure from motion(SFM)technology,the accuracy and efficiency of SFM reconstruction for a different number of image datasets are usually different.Therefore,image quantity's influence on the 3D reconstruction model's accuracy was analysed to provide a basis for image data acquisition and selection in 3D reconstruction application.The image dataset was obtained by shooting target objects with a handheld digital camera,and different numbers of image data sets were selected to perform SFM point cloud and mesh reconstruction respectively.By comparing the model accuracy,noise,and completeness of each 3D model with laser scanning model,the experimental results show a certain correlation between the number of images and the accuracy of the 3D model.The more the number of images,the higher the completeness of the reconstructed model.However,the increase of image quantity will lead to decreased model accuracy and noise level aggravation due to the SFM feature point matching error,registration error,and the impact of redundant scenes in the image.It is necessary to analyse the characteristics and requirements of 3d model production in practical engineering applications and select datasets with appropriate image quantity to perform 3D model reconstruction.