摘要
FPCB即柔性印刷电路板,是一种以柔性绝缘材料为基材制造而成电路板,且有弯曲、卷绕、折叠性好的特点,可以满足电子产品柔性要求,在电子信息产业中得到了广泛应用。而在FPCB柔性印刷电路板生产过程中可能会出现各式各样的缺陷,这些缺陷会对产品有着严重的影响,目前对FPCB的检测主要是以人工目视为主,但是人工检测存在着一些不确定的因素,例如人体的疲劳程度和注意力都对检测准确率有着较深的影响。目前,人工智能、即深度学习在近年来发展逐渐趋于成熟,把深度学习的方法应用到缺陷检测领域是一个有较好前景的研究方向,但是同样也存在着一个问题,工厂良品率的要求使得无法采集到较多的缺陷数据,深度学习方法又需要足够多的数据作为支撑。为了解决这一问题,提出了传统图像处理结合深度学习的方法,并在FPCB数据集上进行验证,利用该方法可以完全满足实际检测需求。
FPCB is a flexible printed circuit board,which is a circuit board made of flexible insulating material as the base material.It has the characteristics of good bending,winding and folding properties and can meet the flexibility requirements of electronic products.It is widely used in the electronic information industry.In the production process of FPCB flexible printed circuit boards,various defects may occur,and these defects will have a serious impact on the product.At present,the inspection of FPCB is mainly based on human eyes,and there are some inconsistencies in manual inspection.Certain factors,such as human fatigue and attention,have a profound impact on the overall detection accuracy.Now artificial intelligence,that is,deep learning,has gradually matured under the development in recent years.It is a promising direction to apply deep learning methods to the field of defect detection,but there is also a problem.The requirements of factory yield make it impossible to obtain more defect data,and the deep learning method needs enough data as the support.In order to solve this problem,this paper proposes a method of traditional image processing combined with deep learning,and validates it on the FPCB dataset.This method can fully meet the actual detection needs.
作者
张鲲
邓明星
ZHANG Kun;DENG Mingxing(College of Automobile and Traffic Engineering,Wuhan University of Science and Technology,Wuhan 430065,China)
出处
《智能计算机与应用》
2023年第7期53-57,63,共6页
Intelligent Computer and Applications
基金
国家自然科学基金(51975426)
武汉市科技计划项目(2019010701011393)
湖北省教育厅科学技术研究项目(B2013238)。
