基于网格标记的视觉坐标测量技术

【作者】 吴焕明；

【导师】 方漪；

【作者基本信息】 青岛大学 ， 计算机软件与理论， 2003， 硕士

【摘要】 本论文对视觉坐标测量技术进行了系统的研究，建立了一个基于网格标记的四目视觉坐标测量系统的模型，并基于这一模型研制了一台四目视觉坐标测量仪，这一仪器用于金属板料变形过程的应变测量，得到了较好的效果。 本论文提出的基于网格标记的四目视觉坐标测量系统的模型首先需要在被测量物体上覆盖一层特殊的网格，把网格点作为测量特征点来处理，通过对带有网格标记的图象对的处理、计算，得到网格点的三维坐标。本论文着重研究了针对带有网格标记的图象这个特殊的处理对象进行图象处理、立体匹配等算法的优化的问题，提出或者改进了一些可行的算法。同时该系统采用四个摄象机作为图象采集源，本论文从理论上对这种特殊的测量模型进行了分析，设计了坐标计算的算法。具体来说研究结果包括以下几个部分： 1．摄象机标定以及空间点三维坐标计算。根据系统的精度要求，摄象机模型的选择的是线性模型。论文提出了基于网格标记的线性摄象机的标定方法，建立了基于线性摄象机模型的立体视觉模型，并设计了四台摄象机匹配进行空间点三维坐标计算的算法。 2．图象采集与处理。设计了硬件独立的图象采集接口。基于网格标记图象，设计了一系列的合理的图象处理步骤，包括图象分割并二值化、二值图象闭运算、细化以及基于B样条插值的亚象素图象处理技术。系统的图象处理算法都是根据网格标记这个特殊的处理对象进行重新设计或者优化设计的。其中独立设计了自适应的移动窗口灰度阈值图象分割法，优化设计了并行模板匹配细化算法，独立提出了基于三次均匀B样条插值放大局部图象的亚象素图象处理算法，并都取得了较好的处理效果。 3．网格点提取以及匹配。基于以前研究者的工作，提出了从细化网格图象中提取网格点的算法。独立的提出了基于网格点相邻关系建立二维表格的网格点匹配算法。 4．坐标数据的后处理。设计了通用的三维显示模块以及后处理数据接口，方便了系统功能的扩展。更多还原

【Abstract】 After systematic researches on computer vision coordinate measurement techniques, this paper proposed a Four-eyed Vision Coordinate Measurement Model (FVCMM) based on Rectangular Grid Mark. A Four-eyed Vision Coordinate Measurement System (FVCMS) was constructed based on this model. It was firstly applied to strain measurement during the process of metal plate deformation and gained some success.In the FVCMM proposed in this paper, a special designed rectangular grid mark is required to be attached to the surface of the object. The grid nodes on the surface are treated as measurement marks. The 3-Dimensional coordinates of these grid nodes are calculated through processing the grid images which captured by four cameras.This paper paid most efforts on the algorithms of image processing and stereo corresponding of the grid nodes concerning the special objects - grid images. Some applicable algorithms are proposed or improved. The FVCMS employed four cameras as image acquisition source. This paper built a theoretical model of this kind of measuring method and designed coordinate calculation algorithm.In summary, the research result described in this paper includes four parts.1. Camera calibration and 3-Dimensional coordinates’ calculation. Regarding the precision requirement of the system, a simple linear camera model is selected and a calibration method based on grid image is proposed. Based on linear camera model, this paper constructed a stereo vision model and designed an algorithm for 3-Dimensional coordinates’ calculation through four cameras.2. Image acquisition and processing. A hardware independent image acquisition interface is designed. Aimed at grid images, a reasonable image processing procedure is employed, including segmentation, binarization, close, thinning and sub-pixel image processing technique based on B-spline interpolation. The algorithms employed in this paper are redesigned or improved for grid images. For example, an adaptable moving window gray scale threshold image segmentation method is proposed originally, a parallel template matching thinning algorithm is improved greatly, a sub-pixel image processing algorithm based on B-spline interpolation is present independently.3. Grid nodes extraction and corresponding. Based on former researcher’s works, an algorithm for grid nodes extraction from thinned grid image is present. A brand new method of grid nodes corresponding is proposed based on 2-Dimensional Neighbor Node Table.4. Coordinate data visualization. A general 3-Dimensional coordinate data visualization model and a general post-processing data interface are designed for extension of the system’s function.更多还原