基于改进的局部表面凸性算法三维点云分割

点云分割是点云分类、识别以及三维重建等处理的基础,分割结果对后续应用影响巨大。本文提出利用连通点集改进局部表面凸性算法中邻近点关系的方法,解决目前激光三维成像系统点云分割算法在处理复杂环境散乱点云时存在分割过度及分割不充分的问题,通过主顶点与周围点构成连通集,作为分割判断局部子点集,形成有效分割区域。该方法解决了常用点云分割方法无法对形状不规则物体进行有效分割的问题,提高了分割精度。算法实验结果表明,相比于最小切割算法和区域生长算法,基于连通点集的改进局部表面凸性算法对实际路面环境信息的分割效果更好,并能在一定程度上避免分割过度和分割不充分的情况,证明该方法适用于复杂环境散乱点云数据分割。     

An entropy-based filtering approach for airborne laser scanning data

Parameter-tuning is a challenging task when generating digital terrain models from airborne laser scanning (light detection and ranging, LiDAR) data. To address this issue, this paper presents a filtering method for near-infrared laser scanning data that exploits the principle of entropy maximization as the optimization objective. The proposed approach generates ground elevation of point cloud by constructing a triangulated irregular network, calculates the entropy of the elevation from different parts, and automatically separates ground and non-ground points by the principle of entropy maximization. Experimental results from different ground surfaces show that the proposed entropy-based filtering method can effectively extract bare-earth points from the point cloud without adjusting thresholds.     

360°连续扫描的便携式三维激光雷达设计与开发

针对三维激光雷达在国内有巨大的应用市场,而国外商业三维激光雷达十分昂贵的现状,开发了一种360°连续扫描的便携式三维激光雷达系统。采用一个二维激光扫描仪与高精度转台连接,通过步进电机控制形成三维激光扫描。在分析系统误差来源的基础上,提出了系统误差校正方法,给出了三维坐标精确计算公式。实验结果表明,系统测量精度高,数据质量好。系统作用距离80 m,测距精度可达6 mm,测量速度每秒7256点,可满足室内室外大规模场景三维数据快速获取的需求,而成本仅是国外同类商业三维激光雷达价格的四分之一左右,且重量轻体积小,携带方便。     

From a Point Cloud to a Simulation Model—Bayesian Segmentation and Entropy Based Uncertainty Estimation for 3D Modelling

The 3D modelling of indoor environments and the generation of process simulations play an important role in factory and assembly planning. In brownfield planning cases, existing data are often outdated and incomplete especially for older plants, which were mostly planned in 2D. Thus, current environment models cannot be generated directly on the basis of existing data and a holistic approach on how to build such a factory model in a highly automated fashion is mostly non-existent. Major steps in generating an environment model of a production plant include data collection, data pre-processing and object identification as well as pose estimation. In this work, we elaborate on a methodical modelling approach, which starts with the digitalization of large-scale indoor environments and ends with the generation of a static environment or simulation model. The object identification step is realized using a Bayesian neural network capable of point cloud segmentation. We elaborate on the impact of the uncertainty information estimated by a Bayesian segmentation framework on the accuracy of the generated environment model. The steps of data collection and point cloud segmentation as well as the resulting model accuracy are evaluated on a real-world data set collected at the assembly line of a large-scale automotive production plant. The Bayesian segmentation network clearly surpasses the performance of the frequentist baseline and allows us to considerably increase the accuracy of the model placement in a simulation scene.     

一种面向空间非合作目标位姿测量应用的三维点云滤波算法

针对激光位姿敏感器获得的原始点云有噪声和直接参与解算消耗星上计算资源过大问题,给出一种适用于空间非合作目标位姿测量的点云滤波和特征提取算法。应用仿真的方法分别验证了算法滤除空间随机噪声和点云降采样的有效性,验证了特征点对目标位姿变化和高斯测量噪声的鲁棒性。在非合作目标绕飞、抵近、捕获全物理试验平台上,以扫描激光位姿敏感器获得的原始点云数据为输入,验证了算法在实际空间目标位姿测量中的性能。试验结果表明,该算法实现了原始点云93.1%的降采样,节省了92.9%的位姿解算时间,可有效提升星上数据处理的效率和姿态解算的实时性。     

Study on fast linear scanning for a new laser scanner

In the field of lidar system design, there is a need for laser scanners that offer fast linear scanning, are small size and have small a rotational inertia moment. Currently, laser scanners do not meet the above needs. A new laser scanner based on two amplified piezoelectric actuators is designed in this paper. The laser scanner has small size, high mechanical resonance frequencies and a small rotational inertia moment. The size of the mirror is 20 mm×15 mm. To achieve fast linear scanning performance, an open-loop controller is designed to compensate the hysteresis behavior and to restrain oscillations that are caused by the mechanical resonances of the scanner's mechanical structure. By comparing measured scanning waveforms, nonlinearities and scan line images between the uncontrolled and controlled scanner, it was found that the scanning linearity of linear scanning was improved The open-loop controlled laser scanner realizes linear scanning at 250 Hz with optical scan angle of ±12 mrad.     

Standard artifact for the geometric verification of terrestrial laser scanning systems

Terrestrial laser scanners are geodetic instruments with applications in areas such as architecture, civil engineering or environment. Although it is common to receive the technical specifications of the systems from their manufacturers, there are not any solutions for data verification in the market available for the users. This work proposes a standard artifact and a methodology to perform, in a simple way, the metrology verification of laser scanners.The artifact is manufactured using aluminium and delrin, materials that make the artifact robust and portable. The system consists of a set of five spheres situated at equal distances to one another, and a set of seven cubes of different sizes. A coordinate measuring machine with sub-millimetre precision is used for calibration purposes under controlled environmental conditions. After its calibration, the artifact can be used for the verification of metrology specifications given by manufacturers of laser scanners.The elements of the artifact are destinated to test different metrological characteristics, such as accuracy, precision and resolution. The distance between centres of the spheres is used to obtain the accuracy data, the standard deviation of the top face of the largest cube is used to establish the precision (repeatability) and the error in the measurement of the cubes provides the resolution value in axes X, Y and Z. Methodology for the evaluation is mainly supported by least squares fitting algorithms developed using Matlab programming.The artifact and methodology proposed were tested using a terrestrial laser scanner Riegl LMSZ-390i at three different ranges (10, 30 and 50 m) and four stepwidths (0.002°, 0.005°, 0.010° and 0.020°), both for horizontal and vertical displacements. Results obtained are in agreement with the accuracy and precision data given by the manufacturer, 6 and 4 mm, respectively. On the other hand, important influences between resolution and range and between resolution and stepwidth are observed. For example, the two smaller cubes cannot be well detected in any case and, as must be expected, the increase in range and stepwidth produces a decrease in the quality of the detection for the larger ones.     

The use of low density high accuracy (LDHA) data for correction of high density low accuracy (HDLA) point cloud

Coordinate measuring techniques rely on computer processing of coordinate values of points gathered from physical surfaces using contact or non-contact methods. Contact measurements are characterized by low density and high accuracy. On the other hand optical methods gather high density data of the whole object in a short time but with accuracy at least one order of magnitude lower than for contact measurements. Thus the drawback of contact methods is low density of data, while for non-contact methods it is low accuracy. In this paper a method for fusion of data from two measurements of fundamentally different nature: high density low accuracy (HDLA) and low density high accuracy (LDHA) is presented to overcome the limitations of both measuring methods. In the proposed method the concept of virtual markers is used to find a representation of pairs of corresponding characteristic points in both sets of data. In each pair the coordinates of the point from contact measurements is treated as a reference for the corresponding point from non-contact measurement. Transformation enabling displacement of characteristic points from optical measurement to their match from contact measurements is determined and applied to the whole point cloud. The efficiency of the proposed algorithm was evaluated by comparison with data from a coordinate measuring machine (CMM). Three surfaces were used for this evaluation: plane, turbine blade and engine cover. For the planar surface the achieved improvement was of around 200 µm. Similar results were obtained for the turbine blade but for the engine cover the improvement was smaller. For both freeform surfaces the improvement was higher for raw data than for data after creation of mesh of triangles.     

基于三维激光扫描的铁路罐车点云优化处理方法

通过对三维激光扫描技术在铁路罐车中的应用研究,对扫描过程中发现的点云残缺及噪点问题进行研究分析,提出一种快速有效点云优化处理方法,新方法包括sp-H点云预处理和Eti-G建模优化。验证试验表明,采用新的点云优化处理方法可以在较短时间内将扫描点云进行去噪及残缺修补,并能够快速高效进行模型重建,提升了不同工况环境下扫描点云的适用性,提高了扫描工作效率和准确度,铁路罐车容积测量结果的相对扩展不确定度达到2.4×10?3,为铁路罐车容积扫描最高准确度等级,为三维激光扫描相关技术的发展提供一定借鉴。     

A performance evaluation test for laser line scanners on CMMs

This paper presents a performance evaluation test for laser line scanners on 3D coordinate measuring machines (CMMs). Laser line scanners are becoming more popular in recent years, mainly for free form inspection tasks and reverse engineering. Error specification of these scanners is difficult because of many influencing factors like surface quality, surface orientation and scan depth. Therefore, procedures for evaluation and verification of conventional contact probes (e.g. touch-trigger probes) are not appropriate for non-contact laser line scanners. A straightforward test method that uses a planar test artefact is proposed. It enables to identify the influence of in-plane and out-of-plane angle, as well as scan depth on systematic and random errors of the laser scanner. Experimental results show that the tested commercial laser scanner, after calibration, exhibits systematic errors of about 10 μm.     

基于机载激光雷达点云的雪道坡度提取算法

滑雪场的雪道坡度信息不仅是雪场规划和建设所需的必要内容,也是滑雪运动员提高成绩、减少运动损伤的重要数据。传统的手动测量方式费时费力,且雪场的低温环境不利于测量人员长时间作业。为了解决滑雪场的雪道坡度计算问题,提出了一种基于激光雷达点云的雪道坡度自动提取算法。使用投影高程差滤波、聚类分割等算法对雪场点云进行预处理,获取雪道边缘特征数据。提出了一种自适应阈值的拟合算法,用于雪道中线的提取,该算法可根据点云数量信息及坐标信息,自动选择阈值拟合雪道中线,消除了实际应用中部分点云数据稀疏或缺失对中线拟合带来的不利影响。数据处理结果表明,该方法计算所得坡度值与多次手动测量平均值之间的相对误差为2.2%。这一方法对激光雷达在雪场目标上的应用、雪场测绘中雪道坡度计算等具有参考意义。     

基于IMU的3D扫描仪倾角标定方法北大核心CSCD

3D扫描仪以精度高、速度快、自动处理、稳定性好等优势在场景三维重建、智能制造、自动驾驶、虚拟现实等领域得到了广泛的应用。目前3D扫描仪设备多是安装在三脚架上,容易因为地面不平使得扫描仪的转轴和重力方向产生夹角,最终导致拍摄出的全景图和点云倾斜,成为影响三维重建精度的关键难题。针对该问题,提出一种借助惯性测量单元标定3D扫描仪倾角的方法。该方法利用惯性测量单元输出的加速度信息,通过一个面阵相机建立惯性测量单元和3D扫描仪的联系,将惯性测量单元的加速度信息转到扫描仪的转轴上。利用该信息计算出扫描仪转轴与重力方向的夹角,并将该倾角补偿到生成的全景图和点云中,校正倾斜模型。通过实验对校正前后计算的角度进行对比,实验结果证明倾角测量误差可以控制在0.5°以内。并且经过标定后,精度至少提升20%,验证了整个标定方法的可靠性和准确性。     

基于视觉SLAM和目标检测的语义地图构建

语义信息对于移动机器人理解环境内容、执行复杂任务至关重要,针对ORB-SLAM2构建的点云过于稀疏、缺乏语义信息、点云所占存储空间大等问题,提出将目标检测算法与视觉SLAM（同时定位与地图构建）技术紧密结合,构建环境的稠密点云语义地图。首先,通过目标检测网络YOLO v3及对象正则化准确获取物体的2D标签,并经过ORB-SLAM2算法构建环境的稀疏点云地图,通过含有2D标签的彩色图像和对应的深度图像以及关键帧来生成含有语义信息的稠密点云标签,使用基于图的分割算法对稠密点云进行分割,再将点云标签与分割后的点云进行融合,进而构建环境的稠密点云语义地图。文中方法在TUM公开数据集上进行试验,实验表明可以构建出效果较好的语义地图。与传统的ORB-SLAM2相比,此系统在构建地图的过程中,相机的绝对位姿误差和绝对轨迹误差分别减少了16.02%和15.86%,提高了建图精度。为了减小点云地图的存储空间,方便移动机器人进行避障和导航,最终将所构建的语义地图转换为八叉树地图。     

DNA microarray scanner with a DVD pick-up head

A low-cost highly sensitive DNA microarray scanner for fluorescent detection is developed based on the pick-up head of a commercially available optical storage device, DVD. A laser beam of 650 nm, generated by a DVD laser diode, is used for dynamic auto-focusing as well as the excitation of Cy5 fluorescent dye. The fluorescence intensity emitted from Cy5 dye is measured by a photomultiplier tube (PMT). In contrast to other microarray scanners, the DVD-based scanner offers the auto-focusing function using the focus error signal (FES) and a voice coil motor (VCM), and this enables fast response, high accuracy and compact size. The fluorescence-detecting performance of the scanner is inspected by using a commercial BAC (bacterial artificial chromosome) oligonucleotide chip and a scanner evaluation microarray (DS01). Experiments have shown that the DVD-based scanner meets the limit of detection, ensuring the feasibility of a low-cost, highly sensitive DNA microarray scanner.     

Low-Cost Laser Scanning System Design

3D data is the main element of three-dimensional models that are commonly used in many different application areas such as architecture, archeology, medicine, film production, reverse engineering, etc. In this study, we discuss designing and utilizing opportunities for a low-cost triangulation-based laser scanner. To evaluate the system, a test object is specially designed, and the accuracy and operation of the system are verified by comparing the data output of the laser scanner with the terrestrial photogrammetric method. The designed system is also verified using different objects. At the end of the study, we calculate the precision of the scanned sample models to be 0.1–0.2 mm and RMS, 0.59 mm.     

三维形貌柔性测量系统标定方法及验证

为了避免机器人模型误差对三维形貌柔性测量系统手眼标定的影响,对手眼关系的标定方法进行了研究。提出了一种融合特征点拟合的手眼标定方法。将三维形貌扫描仪安装在工业机器人末端搭建三维形貌柔性测量系统。标定时,首先利用激光跟踪仪对工业机器人末端法兰盘坐标系进行测量,得到两者转换关系;然后,利用三维形貌扫描仪和激光跟踪仪对空间固定的特征点组进行测量,利用特征点约束和基于罗德里格矩阵的算法求解两者转换关系即可间接地求解出手眼关系。基于ATOS三维扫描仪、安川HP20D机器人和API公司生产的激光跟踪仪进行了手眼标定实验,并进行了精度验证。结果表明:标定后的三维形貌柔性测量系统,其重复性测量精度(3σ)不超过0.1 mm,长度测量精度的均方根误差在0.2 mm以内,点云拼接精度优于±0.7 mm。该方法有效避免了传统手眼标定过程中会引入机器人模型误差的问题,在求解手眼关系解时采用了线性的解法,并且适用于三维形貌柔性测量系统。     

Accuracy improvement of laser line scanning for feature measurements on CMM

Because of its high speed and high detail output, laser line scanning is increasingly included in coordinate metrology applications where its performance can satisfy specified tolerances. Increasing its accuracy will open the possibility to use it in other areas where contact methods are still dominant. Multi-sensor systems allow to select discrete probing or scanning methods to measure part elements. Decision is often based on the principle that tight toleranced elements should be measured by contact methods, while other more loose toleranced elements can be laser scanned. This paper aims to introduce a method for improving the output of a CMM mounted laser line scanner for metrology applications. This improvement is achieved by filtering of the scanner's random error and by combination with widely spread and reliable but slow touch trigger probing. The filtered point cloud is used to estimate the form deviation of the inspected element while few tactile obtained points were used to effectively compensate for errors in the point cloud position.     

基于曲面分割的三维点云物体识别

提出了一种基于局部描述符的三维点云物体识别算法.算法首先得到点云的邻域、法线矢量等相关信息,通过邻域进一步得到形状索引值.特征点的提取以形状索引值为依据,以每个特征点为基点对曲面根据欧式距离和矢量夹角分割.分割的曲面片进行等间距划分为多个欧氏距离同心圆,以特征点切平面为基平面投影,并进行等角度抽样,通过抽样点相对特征点的法线矢量及测地距离变化曲线,建立曲面片的二维描述,从而把三维识别转化为二维.根据算法建立模型数据库,给定一个物体,通过和模型数据库中的曲面描述进行比对,得到潜在的识别结果,再通过迭代最近点算法,得到最终的识别结果.最后,通过大量具体实验验证了算法的有效性,并给出了算法的计算复杂度及耗时对比分析,说明了算法的高效性.     

二维计算全息激光扫描器

本文讨论了计算全息激光扫描器的基本原理,详细推导出平移型二维等间距线阵扫描器的位相函数表达式和平移型二维抛物线扫描器的位相函数表达式。本文还详细介绍了上述二种扫描器的制作方法,并给出实验结果。对于等间距线阵扫描器,作者提出了一种简单易行的制作方法。