基于改进随机Hough变换的快速中心检测方法

A Fast Center Detecting Method Based on Improved Randomized Hough Transform

为了实现芯片焊盘的中心提取,建立了显微视觉系统,研究了焊盘中心坐标的图像检测方法.该方法将改进的随机Hough变换与最小二乘法结合起来,实现对焊盘中心坐标的快速准确定位.首先,根据焊盘基本为圆形的特征,初步确定使用Hough变换法和最小二乘法进行中心检测.其次,分析了这两种方法存在的问题和不足,引出了改进的随机Hough变换法.最后,根据焊盘边缘凹凸不平的特点,将改进的随机Hough变换法和最小二乘法结合起来,通过随机Hough变换检测出焊盘中心的大致坐标,进而锁定焊盘边缘,再通过最小二乘法对中心坐标进行修正而得到精确的亚像素级中心坐标.实验结果表明,该方法的检测精度可达0.57μm,运行时间在12 ms以内,满足显微视觉系统的精度高、速度快和抗干扰能力强等要求.

A micro-vision system was established and the image detection method of center coordinates was studied for realizing the center detection of chip pad.This method can achieve fast and accurate location of the pad＇s center coordinates by the combination of the improved randomized Hough transform and the least squares method.Firstly,according to the basic characteristics of the circular pad,Hough transform and the least squares method were determined to be used in primitive detection.Secondly,the improved randomized Hough transform was introduced after analyzing the existing problems and deficiencies of the two methods.Finally,in terms of the pad＇s jagged edges,the improved randomized Hough transform was combined with the least squares method.The rough coordinates of the pad＇s center were detected by using randomized Hough transform,then the pad＇s edge was located,and the exact sub-pixel center coordinates were eventually achieved after being corrected by the least squares method.Experimental results indicate that the detection precision of the algorithm can reach 0.57 μm,and running time is less than 12 ms.The proposed method can satisfy the requirements for precision,rapid response,and reliability of the micro-vision system.