机构地区: 华南理工大学机械与汽车工程学院
出 处: 《华南理工大学学报(自然科学版)》 2008年第5期135-139,共5页
摘 要: 针对目前自动光学检测系统在进行焊点检测时容易出现缺陷误报和漏报,以及智能化程度不高的问题,提出了一种基于神经网络的检测方法.首先采用一种基于熵的多阈值自动图像分割方法来提取焊点;然后定义焊点图像的一系列特征,并通过实验对特征进行选择;最后建立用于焊点分类的BP神经网络.实验证明,基于神经网络的焊点图像检测方法具有较高的准确率. In order to overcome the error alarming and unintelligence of the automatic optical inspection(AOI) system for the solder joint inspection,a new inspecting method based on neural network is proposed.First,an entropy-based multi-threshold algorithm is adopted to automatically segment the image and to extract the solder joints.Second,a series of features of solder joints is defined and are selected according to the experimental results.Thirdly,a BP neural network is established for the solder joints classification.The high accuracy of the proposed method is verified by experiments.
领 域: [自动化与计算机技术] [自动化与计算机技术]