基于改进RANSAC算法的PCB板定位方法的研究  

Research of Positioning Method in PCB Based on Improved RANSAC Algorithm

作　　者：张舞杰[1] 何广栋 聂新桥 

机构地区：[1]华南理工大学机械与汽车工程学院,广东广州510641

出　　处：《机床与液压》2017年第9期84-87,162,共5页Machine Tool & Hydraulics

基　　金：广东省科技计划项目(2013B010134010;2009B010900016;2012A090100012;2014B090921003);广州市科技计划项目(201604010064);国家科技支撑计划项目(2015BAF20B01)

摘　　要：为了快速精确识别定位Mark点从而确定PCB板连片折板位置,提出一种基于改进RANSAC算法的边缘拟合算法。通过对目标模型的固有特性进行分析,定向筛选样本点,得到估计模型;采用分层抽样法抽取样本点进行残差分析;经过多次迭代后筛选出最符合的模型。对比分析了加权最小二乘法拟合、基于Zernike矩的边缘拟合和基于改进RANSAC算法的边缘拟合3种方法在效率和准确性方面的差异。实验证明:基于改进RANSAC算法的边缘拟合法较其他两种方法具有鲁棒性更好和稳定性更高的特点。In order to positioning the Mark points in PCB to ensure the cutting position of PCB quickly and accurately, an edge extraction algorithm was proposed based on the improved RANSAC. An estimation model was obtained by analyzing the inherent proper-ty of the objective model and filtering sample points directionally. The residual analysis was made by extracting sample points with the stratified sampling method. The best model was picked up after times iterations. The efficiency and precision among the least-squares fitting with weight, Zernike Moments fitting and Improved RANSAC fitting were contrasted and analyzed. The experiments prove that the last method has the better robustness and stability.

关 键 词：随机抽样一致性算法 机器视觉 标志点定位 椭圆拟合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]