激光点云数据处理中基于贝叶斯抽样一致性的模型参数稳健估计方法研究

The development of the techology of laser point cloud post-processing is much slower than that of the hardware of LiDAR, which currently restrict the applications of LiDAR. The estimation of the parameters of a model is often involved in the point cloud processing (e.g. the geometric model of a feature in the fitting of point clouds, the rigid transformation model in point cloud registration, and the mapping model of scan-to-image registration). As the point clouds always contain a plenty of noises,it is a focus of research in point cloud processing worldwide how to implement the robust estimation of the parameters of a model. Based on Baysian theory and RANSAC, this proposal carries out the research on Baysian Sampling Consensus (BaySAC) method using convergence evaluation of hypothesis models, which consists of the following contents: the determination of prior probability based on convergence evaluation of hypothesis models; the update of inlier probability in terms of the correctness of the hypothesis data set; the construction of the cost function of the hypothesis testing using Maxmum Likelihood Function; Multiple-BaySAC algorithm; the robust estimation of the main models in point cloud processing. The research on BaySAC robust estimation method will help solving the problems suffered by the robust estimation of the parameters of the models in point cloud processing, so that promote the development of the technology of LiDAR data post-processing.

激光点云数据后处理技术发展的滞后，是制约激光雷达技术应用的瓶颈。点云数据处理中通常会涉及到模型参数的估计（如点云拟合中的特征几何模型、点云拼接中的坐标转换模型、点云与影像配准中的映射模型等）。由于点云数据中通常包含大量的噪声点，因此如何进行模型参数的稳健估计是当前国内外点云数据处理的研究重点。本课题基于贝叶斯理论和随机抽样一致性（RANSAC）方法，对利用模型收敛度统计检验的贝叶斯抽样一致性稳健估计方法进行研究，该研究的主要内容包括：基于假设模型收敛度统计检验的先验概率确定；顾及假设检验数据点集整体正确性的局内点概率更新；基于极大似然函数的假设检验模型评价函数构建；多模型贝叶斯抽样一致性算法；利用贝叶斯抽样一致性的点云数据处理中主要模型参数稳健估计。贝叶斯抽样一致性稳健估计方法的研究将有助于点云数据处理中模型参数稳健估计问题的解决，从而促进激光雷达数据后处理技术的发展。

激光点云数据后处理技术发展的滞后，是制约激光雷达技术应用的瓶颈。点云数据处理中通常会涉及到模型参数的估计（如点云拟合中的特征几何模型、点云拼接中的坐标转换模型、点云与影像配准中的映射模型等）。由于点云数据中通常包含大量的噪声点，因此如何进行模型参数的稳健估计是当前国内外点云数据处理的研究重点。本课题基于贝叶斯理论和随机抽样一致性（RANSAC）方法，对利用模型收敛度统计检验的贝叶斯抽样一致性稳健估计方法进行了研究：研究了顾及假设检验数据点集整体正确性的局内点概率更新；基于极大似然函数构建了阈值无关的假设检验模型评价函数；研究了多模型贝叶斯抽样一致性算法；基于贝叶斯抽样一致性方法实现了车载激光点云的杆状地物提取、机载LiDAR点云数据分类、基于点云的室内导航要素识别、电力线提取等应用。相关研究成果已发表国际SCI论文9篇，EI检索论文3篇，中文核心期刊论文6篇。通过上述关键技术的突破，有助于点云数据处理中模型参数稳健估计问题的解决，从而促进激光雷达数据后处理技术的发展。