线激光测量系统的开发和点云拼接的实现

发布时间：2018-03-20 21:32

  本文选题：线激光测量　切入点：特征提取　出处：《广东工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：工业检测中,随着被测工件越来越复杂和人们要求测量精度越来越高,传统机器视觉测量因其二维局限性已满足不了精密检测的要求。与此同时,激光测量技术因为鲁棒性强,精确度高,越来越受到人们的欢迎。本文针对实际项目需求,设计和搭建线激光测量系统。主要阐述硬件结构的设计和软件框架的搭建,并对程序设计流程作了分析。由于线激光设备具有扫描盲区,面对较复杂的被测工件时无法一次性获得整个工件的三维数据,有时需要多次在不同视角下进行扫描,然后对三维数据进行拼接以获得整个被测工件模型。鉴于此,本文同时对三维点云拼接算法进行了研究。在点云特征提取中,针对线激光扫描的点云数据的结构特点,提出将三维点云模型映射到二维图像中,并修改图像处理中的Canny算法,对点云模型进行特征潜在区域估算,然后基于曲率信息提取特征点。与传统的三维点云特征提取算法比较,该方法具有更高的执行效率。在点云特征描述中,高维特征算子虽然可以直接解决多视图三维点云数据的配准问题,但它的计算较复杂,对大数量级的点云模型来说,匹配效率将大大降低。本文出于运行效率的考虑,提出基于邻域曲率信息的描述方法,用低维描述算子代替高维描述算子,减少计算时间。在点云模型间对应点匹配中,根据特征描述算子构造评价函数,为每个特征点寻找潜在对应点。由于低维特征描述的信息丰富性不强,每个特征点往往匹配到多个对应点,于是需要作进一步筛选。本文基于刚性变换原则,提出区域投票制度的方式实现对应点的精匹配。本文最后对拼接算法进行了实验和对设计的线激光测量机进行了平整度检测应用测试。首先,测量机对实际物体多次扫描得到不同视角下的点云数据模型,通过本文提出的特征提取算子、特征描述算子和对应点匹配算法对两点云模型进行特征分析。然后,基于迭代最近点算法对两点云模型进行拼接,验证了拼接算法的正确性和有效性。在实际应用测试中,本文对线激光测量机的标定做了实验分析,并利用线激光测量机的平整度检测功能,对几种平整度要求较严格的工件进行实际测量,给出测量结果和工件的3D扫描图形,证明测量机功能的有效性。
[Abstract]:In industrial detection, with the increasing complexity of the workpiece under test and the increasing requirement of measuring precision, the traditional machine vision measurement can not meet the requirement of precision detection because of its two-dimensional limitation. At the same time, the laser measurement technology has strong robustness. This paper designs and builds the line laser measurement system according to the actual project demand, mainly expounds the design of the hardware structure and the construction of the software frame. The program design flow is analyzed. Because the line laser equipment has scanning blind area, the 3D data of the whole workpiece can not be obtained at one time in the face of the more complex workpiece being tested, and sometimes it is necessary to scan the whole workpiece in different angle of view several times. Then the 3D data are spliced to obtain the whole workpiece model. In view of this, the algorithm of 3D point cloud mosaic is also studied in this paper. In the feature extraction of point cloud, the structural characteristics of the point cloud data scanned by line laser are analyzed. The 3D point cloud model is mapped to two-dimensional image, and the Canny algorithm in image processing is modified to estimate the potential feature region of the point cloud model. Then the feature points are extracted based on curvature information. Compared with the traditional 3D point cloud feature extraction algorithm, this method has higher execution efficiency. Although the high dimensional feature operator can directly solve the registration problem of multi-view 3D point cloud data, its calculation is more complicated, and the matching efficiency will be greatly reduced for the point cloud model with large order of magnitude. In this paper, a description method based on neighborhood curvature information is proposed, in which the low dimensional description operator is used instead of the high dimensional description operator to reduce the computing time. In the point cloud model corresponding point matching, the evaluation function is constructed according to the characteristic description operator. Finding potential correspondence points for each feature point. Because of the low dimensional feature description is not rich in information, each feature point often matches to more than one corresponding point, so we need to make further screening. This paper based on rigid transformation principle, In this paper, the method of regional voting system is proposed to realize the precise matching of corresponding points. Finally, experiments are carried out on the splicing algorithm and the flatness detection and application test of the designed line laser measuring machine are carried out. The point cloud data model with different angle of view is obtained by scanning the actual object several times. The feature extraction operator, feature description operator and corresponding point matching algorithm are used to analyze the feature of the two-point cloud model. Two point cloud model is stitched based on iterative nearest point algorithm, which verifies the correctness and validity of the stitching algorithm. In the practical application test, the calibration of the line laser measuring machine is analyzed experimentally. By using the flatness detection function of the line laser measuring machine, the actual measurement of several workpieces with strict flatness requirements is carried out, and the measuring results and 3D scanning figure of the workpiece are given, which proves the validity of the function of the measuring machine.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN249;TP391.41
，

本文编号：1640869