Homography of central points for optical distortion compensation

The paper presents a novel radial lens distortion modelling method for vision systems. It is based on single images of chessboard planar pattern, in which calibration points (corners of chessboard squares) are localised. The points, which are situated near the image centre, where the distortion is negligible, are used to determine undistorted grid of calibration points by the estimation of homographic matrix. The differences in the localisation of undistorted and distorted points are used to find parameters of radial model by the linear least square method. Having the distortion model, the dense compensation is performed with support of bilinear interpolation and a sparse compensation by Newton iterative scheme providing subpixel accuracy.     

贴片电阻表面缺陷自动识别方法

贴片电阻生产过程中的缺陷主要依靠人工在显微镜下检测,速度慢、长期成本高、误检率高.针对贴片电阻单元具有排列整齐、结构简单、图像灰度级少的特点,在贴片电阻图像二值化、边缘提取、直线检测基础上,以相邻电阻单元的相关系数作为电阻缺陷判别依据,提出基于子图投影匹配的快速缺陷检测方法.采用主分量分析法压缩图像数据量,提取缺陷特征,以基于支持向量机对贴片电阻缺陷进行分类并建立实验系统.缺陷检测及识别实验表明,缺陷检测正确率为92.5oo,算法的快速性和识别准确度满足系统快速高精的要求.     

一种新的求解拟单调变分不等式的压缩投影算法

2020年Liu和Yang提出了求解Hilbert空间中拟单调且Lipschitz连续的变分不等式问题的投影算法,简称LYA。本文在欧氏空间中提出了一种新的求解拟单调变分不等式的压缩投影算法,简称NPCA。新算法削弱了LYA中映射的Lipschitz连续性。在映射连续、拟单调且对偶变分不等式解集非空的条件下得到了NPCA所生成点列的聚点是解的结论。当变分不等式的解集还满足一定条件时,得到了NPCA的全局收敛性。数值实验结果表明NPCA所需的迭代步数少于LYA的迭代步数,NPCA在高维拟单调例子中所需的计算机耗时也更少。     

光谱法计算机三维成象与再现系统

本文提出一种全新的计算机立体视觉新方法—光谱法三维成象与再现.文中分析了该方法的基本原理和系统结构,并推导出了获取三维面形信息的三角算法计算公式,最后给出了实验结果.结论表明,该方法作为一种全新的立体视觉方法,具有原理简洁、计算快、精度高的特点,能够对物体进行快速成象并再现物体的三维形状,且结构简单,有很强的实用性.     

基于单幅数字散斑投影及图像相关的离面振动测量

采用单幅数字散斑图投影及高速数字图像采集技术,研究了动态离面位移的测量。采用商用液晶投影仪将计算机产生的模拟散斑图投影到待测动态变形物体表面,由高速数字图像采集设备摄取并保存变形散斑图,采用时间序列数字图像相关软件计算出物体表面各点随时间变化的离面位移。这种方法用于振动分析时不仅可以获得振型分布,而且还可以获得各点的振幅值。与现有的激光频闪照相测振及激光多普勒测振等方法相比,具有光学系统简单,可全场定量测量。悬臂梁振动实验结果证实了该方法的有效性。     

Higher‐Dimensional Unification with continuous and fuzzy coset spaces as extra dimensions

We first review the Coset Space Dimensional Reduction (CSDR) programme and present the best model constructed so far based on the , 10‐dimensional E₈ gauge theory reduced over the nearly‐Kähler manifold with the additional use of the Wilson flux mechanism. Then we present the corresponding programme in the case that the extra dimensions are considered to be fuzzy coset spaces and the best model that has been constructed in this framework too. In both cases the best model appears to be the trinification GUT .     

Linear calibration procedure for the phase-to-height relationship in phase measurement profilometry

Calibration of the relationship between height and phase is of uttermost importance to perform accurate 3D measurements in phase measurement profilometry. This work reports a different approach to this problem by first looking at the analytical expression for this relationship and determining the regime spanned by the fringe analysis method. The conclusions thus ascertained, amply justify confronting the analytical expression with a simple normalization procedure of the experimental data, with a remarkable matching between both results. In light of this, a linear calibration procedure with just one plane is proposed and verified experimentally.     

基于投影算子的正则化最小二乘回归

通过引入经验覆盖数（empirical covering number）和投影算子（projection-operator）,从理论上研究正则化最小二乘回归学习算法.与已有的方法相比,一方面简化了回归分析的过程;另一方面,提高了最小二则回归学习算法的误差收敛阶.即,通过引入投影算子,得到了O（m-1）型的收敛阶,这是统计学习理论中关于泛化误差的最佳逼近阶.