• 信息列表
 

基于信息融合的图像测量技术研究-胡尊凤-2009

2013年06月17日 11:52 胡尊凤 点击:[]

 

 作者姓名  胡尊凤
 中文论文题名  基于信息融合的图像测量技术研究
 外文论文题名  STUDY OF IMAGE MEASURING TECHNOLOGY BASED ON INFORMATION FUSION
 论文提交日期  2009-06-24
 学位年度  2009
 中文论文关键词  图像测量,信息融合,边缘检测,亚像素定位
 外文论文关键词  image measurement, information fusion, edge detection, subpixel edge location

 

中文论文文摘


    基于计算机视觉理论的图像测量技术具有非接触、全视场、成本低等优点,信息融合技术能将不同时间、不同空间、不同类型的传感器信息协调成统一的特征表达。本文将信息融合技术应用于图像测量技术中,建立了多源图像测量系统的数学模型,分析了测量原理并设计了总体方案。
    针对单晶硅制备过程中直径参数的测量问题,文中对图像滤波算法与边缘检测算法进行了仿真研究,根据实验结果选定了合适的图像处理算法;应用信息融合技术中的切尾加权融合算法对几个实际棒状物体的尺寸进行了测量,得出切尾率为0.2-0.3时融合估计值最接近真实值这一结论。
    亚像素定位技术是在经典边缘检测算法的基础上发展起来的,它可以通过软件方式提高测量精度。文章在对当前已有亚像素定位算法进行研究的基础上,提出一种改进的正交Fourier-Mellin矩亚像素边缘定位算法,该算法首先利用Canny算子定位所有可能边缘点的像素级精度位置,然后采用正交Fourier-Mellin矩算子以亚像素级精度进一步定位边缘。实验表明该算法运算时间短,是一种高效的亚像素边缘定位算法。
    照明光源的设计也是图像测量技术中的关键问题。文中讨论了图像测量中非相干光照明的几个影响因素,包括照明方式、光源照度、光源质量、待测物表面特征等。构建实验系统对几种不同表面特性的对象进行了测量,对不同照度下对象的测量结果与图像直方图进行了分析,得到图像测量系统中光源的配置原则,最后对测量的误差来源进行了分析。

 

外文论文文摘


    Image measuring technology based on computer vision theory has such advantages as non-contact, full viewing field and low cost. And information fusion technology could coordinate several sensors’ information from different time, space or type to get characteristics of a unified expression. This paper makes use of information fusion technology in image measuring technology, and constructs mathematical model of multi-source image measuring system. Measuring principal is analyzed and core measuring scheme is designed in detail.
    For diameter parameter measuring problem of single-crystal silicon during its manufacturing procedure, image filtering algorithm and edge detection algorithm are respectively simulated and experimented. More proper processing algorithm is chosen according to experimenting results and specific applying object. Then several objects are measured utilizing trimmed mean-based weighed fusion algorithm in information fusion technology. Experiments results show fusion estimation value is closest to real value when trimmed ratio ranges between 0.2 and 0.3.
    Subpixel location technology is developed from classical edge location algorithm, achieving effectively improving measuring accuracy at low cost relatively. Existing subpixel edge detection operators are introduced in this paper, and then an improved Orthogonal Fourier-Mellin Moment subpixel edge location algorithm is presented. This algorithm firstly makes use Canny operator to extract all possible edge points at pixel level and then utilizes subpixel-level edge detection operator, Orthogonal Fourier-Mellin Moment operator to location subpixel level edge location. Experiment results show that this algorithm could achieve equal location accuracy compared with Zernike and OFMM subpixel edge location operator but consume less operating time. It can be concluded that such location algorithm is efficient.
    Illumination design is an essential issue in vision measuring technology. Several measuring-influencing factors of non-coherent lighting illumination are discussed in the paper, consisting of illumination mode, illumination level, lighting quality and surface characteristic of object to be measured. Then measuring system is constructed to test different characteristics of different lighting mode, and measuring results of different illumination level and image histogram are analyzed to get system lighting illumination scheme. System error sources are analyzed at last.
    

 

 

上一条:两性磷酸酯表面活性剂的制备及其在皮革加脂中的应用-安华瑞-2006 下一条:近代服饰虚拟博物馆的研究-齐欢-2009

关闭