基于大数据分析的激光雷达三维图像重建方法  

Lidar 3D image reconstruction method based on big data analysis

作　　者：郝娟 孙皓月 唐振禹 穆莹雪 贾丽坤[3] HAO Juan;SUN Haoyue;TANG Zhenyu;MU Yingxue;JIA Likun(College of Information Engineering,Hebei University of Architecture,Zhangjiakou Hebei 075000,China;School of microelectronics,Xi’an university of electronic science and technology,Xian 710000;Department of Information Management,Hebei University of Architecture,Zhangjiakou Hebei 075000,China)

机构地区：[1]河北建筑工程学院信息工程学院,河北张家口075000 [2]西安电子科技大学微电子学院,西安710000 [3]河北建筑工程学院信息管理系,河北张家口075000

出　　处：《激光杂志》2022年第12期93-97,共5页Laser Journal

基　　金：河北省省属高等学校基本科研业务费研究项目(No.2021QNJS12,2021QNJS04)。

摘　　要：为了获得分辨率高、清晰的激光雷达三维图像,设计基于大数据分析的激光雷达三维图像重建方法。利用K-means聚类算法聚类激光雷达点云数据,获得协方差矩阵的维数;通过高斯过程回归对雷达图像深度建模,得到差值点对应的估计值,通过该估计值实现点云数据内插,强化雷达激光点云数据;使用马尔科夫随机场法构建全局标记优化模型,利用该模型在激光雷达图像像素点与三维距离点之间搭建一个映射关系,获取激光雷达数据结构,由点云数据空间统计模型完成高分辨率的激光雷达三维图像重建。实验结果表明,该方法重建后图像分辨率高,重建分辨率高达1 900,噪声干扰较少且三维图像清晰度高。In order to obtain high-resolution and clear lidar 3D images, a lidar 3D image reconstruction method based on big data analysis is designed. The K-means clustering algorithm is used to cluster LIDAR point cloud data to obtain the dimension of covariance matrix. The radar image depth is modeled by Gaussian process regression and the estimated value corresponding to the difference point is obtained. Through the estimated value, the point cloud data interpolation is realized to strengthen the radar laser point cloud data. The Markov random field method is used to construct the global marking optimization model, which is used to build a mapping relationship between the lidar image pixels and the three-dimensional distance points, obtain the lidar data structure, and complete the high-resolution lidar three-dimensional image reconstruction by the point cloud data spatial statistical model. Experimental results show that this method has high image resolution, reconstruction resolution up to 1 900, less noise interference and three-dimensional image clarity.

关 键 词：大数据 激光雷达 三维图像 高斯回归模型 K-MEANS聚类 

分 类 号：TN209[电子电信—物理电子学]