摘要
由于虚拟现实场景散乱点云图像拼接存在数据冗余与离群点问题,导致点云图像中冗余信息和干扰的离群数据点影响了拼接的准确性。同时,由于不同采集设备和视角造成的坐标差异和位置偏差,导致拼接过程中出现错位、重叠或缺失,进一步影响拼接质量。因此,提出虚拟现实场景散乱点云图像快速拼接算法。结合K邻域算法与z-score标准化处理方法,对点云图像内的干扰离群数据点剔除。引入基于PCA算法的特征向量计算方法,将多张点云图像对齐至同一个坐标系中,避免后续图形拼接过程中错位问题的发生。令点云采集设备以特定方向对目标物实施连续点云图像拍摄,通过拼接空间内点云图像的绕矩推导,实现虚拟现实场景内点云图像的快速拼接。实验表明,所提方法在保持拍摄对象结构真实性的同时,有效提升了点云图像的整体质量,并实现了较为完整的图像拼接效果。
Due to the problems of data redundancy and outliers in the mosaic of scattered point cloud images in virtual reality scenes,redundant information and disturbing outliers in the point cloud images affect the accuracy of the mosaic.At the same time,due to the coordinate differences and positional deviations caused by different acquisition devices and perspectives,misaligment,overlap or missing information may occur during the splicing process,further affecting the splicing quality.Therefore,a fast mosaic algorithm for scattered point cloud images in virtual reality scenes is proposed.The K-neighborhood algorithm and z-score standardized processing method are combined to eliminate the interference outliers in the point cloud image.The feature vector calculation method based on PCA algorithm is introduced to align multiple point cloud images to the same coordinate system,so as to avoid the occurrence of dislocation in the subsequent graphic stitching process.Make the point cloud acquisition equipment take continuous point cloud images of the target in a specific direction,and realize the rapid point cloud image mosaic in the virtual reality scene through the derivation of the point cloud image's rotation moment in the mosaic space.Experiments show that the proposed method can effectively improve the overall quality of the point cloud image and achieve a more complete image mosaic effect while maintaining the structural authenticity of the subject.
作者
杨琦
徐子卿
YANG Qi;XU Zi-qing(Tianhua College,Shanghai Normal University,Shanghai 201815,China)
出处
《计算机仿真》
2025年第2期366-370,共5页
Computer Simulation
基金
上海嘉定职业教育集团2023-2025年职业教育改革和建设项目(JDZJJT-2023-006)。
关键词
虚拟现实场景
离群点剔除图像拼接
绕矩拼接
Virtual reality scene
Outlier rejection
Image splicing
Moment-winding splicing
