摘要
基板引脚图像识别是自动引线键合的关键技术之一。文章针对引线键合过程中基板引脚图像的特点,提出了一种基于神经网络的识别方法:首先从图像中提取了由5个特征值组成的识别特征向量;然后用神经网络设计了分类器,并进行了有效的训练;最后,经过训练的神经网络分类器可对引脚图像进行有效的识别。研究结果表明:该方法具有简单、快速、有效的特点,在少量的训练情况下可以达到很好的识别效果。
The image recognition of substrate draw-foot is a key technique for auto wire bonding.According to the characteristic of substrate draw-foot image,a new pattern recognition method based on neural network is discussed in this paper.Firstly,eigenvectors made up of5eigenvalues are extracted from some substrate draw-foot sample images,and then the sample eigenvectors are used to train a neural network classifier.Finally,the trained classifier can be used to make a judgment of the input eigenvectors effectively.The research results show that this method is simple,rapid and effective even given a few training.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第12期202-204,共3页
Computer Engineering and Applications
基金
国家自然科学基金资助(编号:50390064)
