基于分裂合并的多模型拟合方法在点云分割中的应用

本文基于机器视觉探讨数字摄影测量三维构像下的智能数据处理要素之二:海量点云分割处理技术。多模型拟合方法通过将点云拟合到不同模型中,依照点云空间分布特征和几何结构特征进行分割。针对点云数据量巨大、分布不均匀、结构复杂等特性,本文提出一种基于多模型拟合的点云分割方法。首先通过降采样,采用基于密度分布的聚类方法,实现对点云的预分割。在预分割基础上,利用基于分裂合并的多模型拟合方法对点云进行后续拟合分割。针对平面和弧面,本文采用不同的拟合方式,最终实现对室内密集点云分割。试验结果表明,该方法能够在无须提前设置模型数目的情况下实现点云的自动分割。且相较于现有的点云分割技术,此方法相较于现今的常规方法能取得更好的分割效果,在分割的正确率上要高于现有的常规分割方法,在处理相同数据量的点云分割时,能够达到远低于常规方法的时间消耗。通过本文提出的三维点云分割方法能够实现将大规模、复杂三维点云数据分割为较为精细、具有准确模型参数的三维几何图元,为后续实现大规模、复杂场景的精确三维构象提供有力支持。

一种RANSAC多模型拟合的隧道点云滤波算法

针对目前大多数点云滤波方法的不足,提出基于随机抽样一致(random sample consensus, RANSAC)多模型拟合的隧道点云滤波算法。与单模型拟合滤波方法不同,该算法在每个隧道横断面上使用多个圆模型进行分段拟合,多个模型具有不同的模型参数,将远离所拟合的各圆模型的点作为噪声点进行剔除。实验结果表明,相较于单模型拟合滤波方法,所提出的多模型法具有更低的一类误差和相近的二类误差,对隧道三维激光点云的滤波效果更好。

Simulation of Particle Growth in Gas Phase Polymerization of Butadiene by the Multigrain Model

Gas phase polymerization of butadiene by neodymium catalyst was modeled. The effects of mass and heat transfer resistances in the external boundary layer and within particles, sorption of butadiene in polybutadiene,and deactivation of active sites on polymer particle growth and morphology were studied. Simulation results show that the effects of intraparticle mass and heat transfer resistances on the growth rate of polymer particles are insignificant, and that there is no significant effect of mass transfer resistance on the morphology of polymer particles.The simulation results were compared with the experimental results.

Preliminary Analysis of Non-tidal Variation in the Baijiatuan Gravity Station,Beijing and the Relevance of Gravity Changes and Water Storage

The non-tidal variation gained from continuous gravity observations in stations usually reflects the regional continuous gravity changes. In this paper we focus on studying the non-tidal variation of Baijiatuan station,Beijing where there are two different gravimeters( namely,L&R-804 and PET-031). Based on the original raw tidal records of two gravimeters from 2008 to 2011,we first remove various interference from raw data by the standard procedure software-Tsoft; then we model the solid earth tides, ocean tidal loading and pole tide through related parameters; after that we adopt a new segmented polynomial fitting method based on Tsoft to fit the complex drift of spring gravimeter; and finally we calculate the atmospheric loading effects by a linear regression model. After a series of processing we gain the non-tidal variation of the two gravimeters at Baijiatuan site,Beijing. Furthermore,to analyze the non-tidal variation preliminarily,we study the main component of related tidal data by power spectral density. Comparing the non-tidal variation of two different gravimeters,we find seasonal fluctuations in non-tidal results, which are in accordance with the water storage change. Therefore,we take into account the relevance of gravity changes and water storage based on the gravity data of GRACE and water data of the CMAP model from 2003 to 2011 at different sites in the Chinese mainland( Beijing, Chengdu, Shenyang and Shiquanhe), and make a preliminary analysis on the relationship between gravity changes and water storage.

分裂合并运动分割的多运动视觉里程计方法

目的视觉里程计(visual odometry,VO)仅需要普通相机即可实现精度可观的自主定位,已经成为计算机视觉和机器人领域的研究热点,但是当前研究及应用大多基于场景为静态的假设,即场景中只有相机运动这一个运动模型,无法处理多个运动模型,因此本文提出一种基于分裂合并运动分割的多运动视觉里程计方法,获得场景中除相机运动外多个运动目标的运动状态。方法基于传统的视觉里程计框架,引入多模型拟合的方法分割出动态场景中的多个运动模型,采用RANSAC(random sample consensus)方法估计出多个运动模型的运动参数实例;接着将相机运动信息以及各个运动目标的运动信息转换到统一的坐标系中,获得相机的视觉里程计结果,以及场景中各个运动目标对应各个时刻的位姿信息;最后采用局部窗口光束法平差直接对相机的姿态以及计算出来的相机相对于各个运动目标的姿态进行校正,利用相机运动模型的内点和各个时刻获得的相机相对于运动目标的运动参数,对多个运动模型的轨迹进行优化。结果本文所构建的连续帧运动分割方法能够达到较好的分割结果,具有较好的鲁棒性,连续帧的分割精度均能达到近100%,充分保证后续估计各个运动模型参数的准确性。本文方法不仅能够有效估计出相机的位姿,还能估计出场景中存在的显著移动目标的位姿,在各个分段路径中相机自定位与移动目标的定位结果位置平均误差均小于6%。结论本文方法能够同时分割出动态场景中的相机自身运动模型和不同运动的动态物体运动模型,进而同时估计出相机和各个动态物体的绝对运动轨迹,构建出多运动视觉里程计过程。

Species Abundance Distribution Patterns of a Toona ciliata Community in Xingdoushan Nature Reserve

With the goal of model fitting species abundance distribution patterns of the tree,shrub and herb layers of the natural Toona ciliata community in Xingdoushan Nature Reserve,Enshi Autonomous Prefecture,Hubei Province,we used the data collected from the field survey and employed different ecological niche models.The models tested were the broken stick model(BSM),the overlapping niche model(ONM)and the niche preemption model(NPM),as well as three statistic models,the log-series distribution model(LSD),the log-normal distribution model(LND)and the Weibull distribution model(WDM).To determine the fitted model most suitable to each layer,the fitting effects were judged by criteria of the lowest value of Akaike Information Criterion(AIC),Chi-square and the K-S values with no significant difference(P>0.05)between the theoretical predictions and observed species abundance distribution values.The result showed:(1)The fitting suitability and goodness of fit of the tree,shrub and herb layers by using the three ecological niche models were ranked as:NPM>BSM>ONM.Of the three statistical models,by accepting the fitting results of the three layers,WDM was the best fitting model,followed by LND.By rejecting the fitting tests of the herb layer,LSD had the worst fitting effect.The goodness of the statistical models was ranked as:WDM>LND>LSD.In general,the statistical models had better fitting results than the ecological models.(2)T.ciliata was the dominant species of the tree layer.The species richness and diversity of the herb layer were much higher than those of either the tree layer or the shrub layer.The species richness and diversity of the shrub layer were slightly higher than those of the tree layer.The community evenness accorded to the following order:herb>shrub>tree.Considering the fitting results of the different layers,different ecological niche models or statistical models with optimal goodness of fit and ecological significance can be given priority to in studying the species abundance distribution patterns of T.ciliata communities.

Numerical modeling of simultaneous hydraulic fracturing in the mode of multi-well pads

In order to investigate propagation regularity of hydraulic fractures in the mode of multi-well pads, numerical modeling of simultaneous hydraulic fracturing of multiple wells was conducted. The mathematical model was established coupling rock deformation with fluid flow in the fractures and wellbores. And then the model was solved by displacement discontinuity method coupling with implicit level set method. The implicit method was based on fracture tip asymptotical solution and used to determine fracture growth length. Simulation results showed that when multiple wells were fractured simultaneously, adjacent fractures might propagate towards each other, showing an effect of attraction other than repulsion. Fracture spacing and well spacing had significant influence on the propagation path and geometry of multiple fractures. Furthermore, when multiple wells were fractured simultaneously, stress reversal regions had a large area, and stress reversal regions were distributed not only in the area between fractures but also on the outside of them. The area of stress reversal regions was related to fracture spacing and well spacing. Results indicated that multi-well fracturing induced larger area of stress reversal regions than one-well fracturing, which was beneficial to generating complex fracture network in unconventional reservoirs.