Recognition of multibreak patterns by 8-neighborhood-based General Hough Transform

In this paper, an 8-neighborhood-based Generalized Hough Transform (ENGHT) for the recognition of multibreak patterns is proposed, which is different from the Generalized Hough Transform (GHT). The difference between ENGHT and GHT appears in the process for Reference-table (R-table) configuration and for pattern recognition. The theoretic and experimental results show that ENGHT can be employed in the recognition of multibreak patterns with a high precision and a high recognition speed, and reduce the difficulty during the recognition that arose from the broken patterns to be processed by the conventional GHT. ENGHT can be employed in pattern recognition, especially in the multibreak pattern recognition.     

基于CCD的无接触式的物体测量

基于CCD的物体测量平台，首先对物体图像进行预处理，继而利用Canny算子提取物体边缘，再通过本文改进的Hough变换实现对被测物体边界直线检测和重构，最后通过多项式插值算法实现物体像素尺寸到物体实际尺寸的转换．实验证明，此方法能够实现无接触式物体尺寸较高精度的测量．     

3D particle positioning by using the Fraunhofer criterion

In particle tracking velocimetry, the necessary information is the 3D location of a given particle in space. This information can be obtained by examining the real image or by analyzing the interference fringe recorded on a digital camera. In this work, we measure the three-dimensional position of spherical particles by calculating the Central Spot Size of the interference pattern of a particle diffraction image. The Central Spot Size is obtained by combining the Continuous Wavelet transform and circle Hough transform. The Continuous Wavelet transform allow us in only one step enhanced quality of particle images and sets a threshold to select properly places where a Central Spot Size appear in order to determine its size via the Hough transform. The size and centroid of the Central Spot Size render z and x-y position of a particle image, respectively. The Central Spot Size is related to a criterion of a simplified theory given by the Fraunhofer theory in order to obtain z particle position. Our approach has been applied to simulated and experimental particle images. Simulated particle images show good agreement between actual and calculated Central Spot Size. An average relative error of 0.5% and 1.12% for x-y and z directions, respectively, was found in the analysis. Our experimental particle images were obtained from particle motion inside a channel. The quality of the particle images determines the accuracy of the calculation of the Central Spot Size of a particle image.     

用Hough变换的方法提取图像拐点

拐点特征是模式识别中经常用到的一类不变量．本文利用Ｈｏｕｇｈ变换找出图像轮廓中的直线段，确定其直线方程，将相邻两直线段延长相交得到的交点定义为拐点．由于Ｈｏｕｇｈ变换是统计平均的结果，所以求得的拐点较为稳固，抗噪声性能较好．文中也给出了拐点的具体算法．     

一种正六边形检测方法的研究

提出了一种基于几何特征的正六边形检测方法。在简述Hough变换原理和正六边形特点的基础上,把正六边形的检测分解成直线和圆的检测,以及线段长度和直线夹角的计算。利用检测到的直线,在图像平面内建立直线方程,依据直线方程,计算相邻两条直线的夹角和交点,求出每条边的长度,同时检测6个交点是否均匀分布在同一圆周上。在螺栓头部正六边形检测实验中,角度测量误差为0.38%,边长测量误差为0.85%。结果表明,该方法是有效的,能准确地检测出图像中的正六边形。     

一种改进的RGB Hough车牌校正定位算法

车牌的定位对车牌识别的成功率起到决定性作用。为了通过校正倾斜的车牌提高车牌在倾斜情况下字符识别的准确率,针对小型车蓝底白字的车牌提出了一种基于RGB色彩空间与Hough变换的车牌校正定位算法。对无倾斜情况的车牌可以直接定位,对于存在明显倾斜的车牌可以实现倾斜校正并定位。对车牌无倾斜情况的定位准确率为93%,倾斜车牌的定位准确率为88%。     

An image-based LCD positioning system utilizing the modified FHT method

In this paper, a LCD positioning system based on an image processing method is described. This method is effective in finding the angle of the geometric features in the cross registration mask. However, when two marks appear in the same frame, the problem of situation verification becomes more complex. We use an algorithm that assists with calculating the location and rotation angle for two marks precisely and quickly using the Fast Hough Transform (FHT). This method determines the locations of two masks and calculates the rotation angle from a blurred edge to adjust the LCD position and orientation.     

Extraction of Input Objects from Optical Hough Transform Data

We propose a method of extracting the true length of a line and an ellipse with two axes from the width of envelopes in the parameter domain. We performed computer simulations and, using a Hough transform computer generated hologram filter, also performed optical experiments. Our experimental results were very similar to those of the simulation.     

基于多模板子空间的支持向量机多类自然形状识别

自然物体的检测与识别是机器视觉以及模式识别的重要任务。由于自然物体形状的多样性与柔性以及视觉判别的复杂性,使基于计算机的自然形状物体的准确检测与识别变得比较困难。提出了基于多模板子空间的支持向量机(SVM)多类自然形状识别方法。利用广义Hough变换表示自然形状物体轮廓,针对每个类别通过训练得到多个匹配模板;检测时利用多模板最近邻相关匹配进行粗检测,使用支持向量机进行分类。在相关匹配限定的子空间内收集训练样本,有效地降低了训练样本数目。实验结果证明所提出的自然形状检测与识别方法是十分有效的,大大改进了经典检测算法的检测效果以及自动化程度。     

一种基于单目视觉的移动机器人室内导航方法

针对室内导航的环境特点,提出了一种简单快速的、以踢脚线为参考目标的移动机器人室内导航方法。该方法从图像中提取踢脚线作为参考直线,通过两条直线在图像中的成像特征,提取角度和横向偏离距离作为移动机器人的状态控制输入,从而实现移动机器人的横向运动控制。该方法无需进行摄像机的外部参数标定,大大简化了计算过程,提高了视觉导航的实时性。