【作者】 梁声；

【导师】 邢渊；

【作者基本信息】 上海交通大学 ， 材料加工工程， 2008， 硕士

【摘要】 综合近景摄影测量技术和光栅投影测量技术的实物数据获取方法是目前众多反向工程测量技术中针对大型的、结构复杂的测量对象的最高效的方法之一。这种方法由近景摄影测量获取散布在被测物体上或周围的人工标记点群的三维坐标,再以这些坐标数据作为光栅投影分片测量点云拼接的依据,从而获取得到整体测量数据。这种综合方法既具有光栅投影测量的高效率又消除了数据拼接时的累积误差。该方法需要在被测物体表面贴两类测量标记点,一类是近景摄影测量系统和光栅测量系统都可以识别的,没有固定编号的标记点,称为非编号标记点,此类标记点只需识别其特征的中心位置;另一类是只有近景摄影测量系统可以识别的、表面特征各为不同并且有固定数字编号的标记点,称为编号标记点,这类标记点在图像识别中既需要识别其特征的中心位置,又需要识别其具体的编号代码。其中,非编号标记点的作用是作为每次光栅测量得到的点云拼合的参照。而编号标记点的作用则是通过近景摄影测量原理,根据其自身的空间坐标和两类测量标记点之间的距离关系,将不同的数字图像上的非编号标记点进行匹配,以获得贴放在物体表面上的非编号标记点的空间坐标信息,从而提高点云拼合的精度。全文按照摄影测量的实现过程主要研究了以下内容:1.本文在MATLAB平台上首先利用一系列的图像处理技术将原图像转化为含有较少干扰杂质块的二值图像。2.利用Zernike矩不变量和支持向量机(SVM)方法实现了编号标记点的识别。采用Hamming距离来求取识别编号标记点所需要的Zernike特征向量的维数。3.对标记点的粘贴方式进行约定,基于空间拓扑关系划分了Delaunay三角网格,建立了每幅像片上非编号标记点匹配关系,对非编号标记点的匹配精确。4.采用DLT解法计算图像的投影矩阵和标记点的三维坐标。对超定方程组采用最小二乘解法计算。更多还原

【Abstract】 For recent measurement which involves large volume, complicated structures and free form surfaces targets, the combination of close range photogrammetry and structured light grating projection measuring is the most fast and efficient method. This combination method features with both the high efficiency of structured light method and the elimination of cumulative errors profited from photogrammetry. The photogrammetry establishes a coordinate group of uncoded points which forms the reference set for merging point-clouds acquired by the structured light measurement method.So, the optics measuring system requires two kinds of artificial points to be pasted to the surfaces of the measured objects. One kind of artificial points can be detected by both photogrammetry and structured light method, and there are no fixed code on them, which are called uncoded points and the center positions of uncoded points are need to be specified. The other kind of artificial points have fixed codes and can be detected by photogrammetry only, and their surface patterns are different from each other, which are called coded points. Among them, the function of the uncoded points is the reference of sewing the point-clouds between different images, and the function of the coded points is to match the uncoded points in images and acquire the space information of them.Along with the process of photogrammetry, the range of this research is showed as the following:1. In the part of image processing, this thesis utilizes a series of image processing techniques to turn the original images into binary images with a little interference using MATLAB Image Processing Toolbox (IPT)2. The thesis recognizes coded points by SVM (Support Vector Machine) method with Zernike moments feature vectors extracted from the coded points in images. The method is based on the Hamming distance to determine the dimensions of Zernike moments feature vectors.3. The regulation for deployment uncoded points was prescribed and The Delaunay triangle grid of each image was generated based on the space topology. The matching result is precise based on invariable Delaunay triangle grid. 4. DLT method was used to calculate the projection matrices of images and 3D coordinates of uncoded points. Overdetermined Linear Equations are solved by least squares method.更多还原