摘要
在手机屏幕的制作过程中,由于作业环境和技术等因素的影响,产品总会出现一些不可避免的缺陷,如气泡、划痕、锡灰等。传统人工和计算机视觉手机屏募缺陷检测方法存在检测能力弱、准确度低、成本大等缺点。为提升手机屏幕缺陷检测准确率,文章在Faster R-cnn模型中以不同方式加入注意力机制CBAM,实验结果表明,改进后模型检测准确度达89.78%,比原模型提高了3.33%.
In the production process of mobile phone screen,due to factors such as operating environment and technology,some products will always have some inevitable defects,such as bubbles,scratches,tin dust,etc.The traditional manual and computer vision methods for mobile phone screen defect detection have some shortcomings,such as weak detection capability,low accuracy and high cost.In order to improve the accuracy of mobile phone screen defect detection,this paper adds attention mechanism CBAM in different ways on the basis of Faster R-enn model.The experimental results show that the detection accuracy of the improved model is 89.78%,which is 3.33% higher than the original model.
作者
查云威
陈志豪
李伟朝
ZHA Yunwei;CHEN Zhihao;LI Weichao(Guangdong University of Technology,Guangzhou 510006,China)
出处
《计算机应用文摘》
2022年第22期78-80,共3页
Chinese Journal of Computer Application