基于马尔科夫随机场的散乱点云全局特征提取

Global Feature Extraction from Scattered Point Clouds Based on Markov Random Field

为了精确提取点云数据中的特征信息,针对激光扫描获取的三维散乱点云数据,提出一种基于马尔科夫随机场(Markov random field,MRF)的散乱点云特征提取方法.首先,根据散乱点的曲率估计及阈值初始化点标号并判定稳定点,将稳定点标记存储在数组中;然后,将优化不稳定点的标号问题转化为随机场标号的能量函数问题,引用贝叶斯估计求后验概率分布函数及MAP-MRF(Maximum a posteriori-Markov random field)框架归约得到目标函数;最后,根据图割法α-expansion算法,利用标号调整过程中标号集相对能量变化得到不稳定点的最优标号集,将其与存储稳定点的数组综合,根据点标号提取特征点.实验结果表明,该方法简单、高效、无需人工调参,能够依据全局能量的变化自适应提取特征,特征提取结果令人满意.

In order to extract the feature information of point clouds data accurately, a new method of feature extraction based on Markov random field （MRF） is proposed. First based on scattered point of curvature estimation and the threshold to initialize labels and determine the stability, the stable point marks stored in the array. Second, the problem of optimal unstable label transform to energy function of the label of the airport. By citing Bayesian estimation for posterior probability distribution function and the MAP-MRF （Maximum a posteriori-Markov random field） reduction, objective function is obtained. Finally according to the graph cut algorithm, using label adjustment process label set relative energy changes get optimal labels of unstable set, and the stable storage array synthesis, by labels rapidly extracts feature points. Experimental results show that the proposed method is simple and fast, and does not need manual setting of threshold. According to the change of global energy, the optimal labeling and feature points are extracted.