基于改进Hough变换的印刷电路板定位配准  

Auto-location and Registration of Printed Circuit Board Based on Improved Hough Transform

作　　者：李孟[1] 姚剑敏[1] 郭太良[1] 

机构地区：[1]福州大学物理与信息工程学院,福建福州350002

出　　处：《电视技术》2015年第9期61-65,70,共6页Video Engineering

基　　金：国家"863"重大专项(2012AA03A301;2013AA030601);国家自然科学基金项目(61101169;61106053);福建省自然科学基金项目(2011J01347)

摘　　要：电路板自动光学检测中常通过检测电路板的边缘直线来确定其旋转角度和平移量从而完成定位配准,存在运算量大、精度不高的问题。对传统Hough变换检测直线方法进行改进,从图像金字塔、约束投票角度、约束搜索点的区域3个方面提高Hough变换的效率,并利用最小二乘拟合法提高检测精度。定位配准时,先运用改进的Hough变换求取电路板边缘直线的近似参数;然后在原图上对边缘直线进行拟合求取电路板的精确旋转角度和平移量,完成定位配准。实验表明本文方法能准确完成电路板的定位配准,对大小为1 920×1 080像素的电路板图像,定位时间约为60 ms,定位误差小于1个像素。Detecting the edge lines of Printed Circuit Board（PCB） to obtain the rotation angle and translation is often adapted in PCB auto optics inspection to locate and registrar PCB image. It has defects in calculation and precision. Image pyramid, voting an- gle constraint and searching area constraint are used to improve the efficiency of Hough transform. Further more, the least squares fitting is adapted to achieve higher accuracy. To complete location and registration of the tested PCB, the improved Hough transform is firstly applied to detect approximate edge lines, and then these edge lines are fitted to get the precise rotation and translation of the PCB. The experimental results show that the proposed method can locate PCB accurately, for a PCB image of 1920 × 1080 pixels it costs about 60 millisecond and the deviation is less than one pixel.

关 键 词：机器视觉 印刷电路板 HOUGH变换 梯度角 

分 类 号：TN911.73[电子电信—通信与信息系统] N391.4[电子电信—信息与通信工程]