微小零件视觉检测系统  

Visual inspection system for tiny parts

作　　者：丁明[1] 黄传辉[1] 

机构地区：[1]徐州工程学院,江苏徐州221008

出　　处：《扬州大学学报（自然科学版）》2011年第3期61-65,共5页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(50975276)

摘　　要：通过提取芯片图像中暗点、边缘、块数、面积和亮点等5种与芯片位置无关而且相互独立、易于提取的图像特征,建立正态分布模型,利用新识别出的损坏或缺陷芯片自动修正模型参数,提高模型的准确度.基于最小风险贝叶斯模式识别构造出各种损坏和缺陷芯片的分类器,对污损、烧蚀、碎裂和电极缺失芯片的正确识别率可以达到90%以上.Based on five kinds of image feature such as dark pixel number,edge pixel number,blocks number,area and bright pixel number which are extracted from chip image irrelevant to chip position,the normal distribution model is established.Newly identified auto-modified model parameters for damaged or defective chips are employed to promote its accuracy.The classifier for various damaged and defective chips based on the minimum risk Bayesian classifier is established to identify stained,ablated,disintegrated and electrode-lost chips with accuracy higher than 90%.

关 键 词：微小零件 损坏芯片 缺陷芯片 模式识别 

分 类 号：TN307[电子电信—物理电子学] TP391.41[自动化与计算机技术—计算机应用技术]