基于改进Faster RCNN的马克杯缺陷检测方法  

Mug Defect Detection Method Based on Improved Faster RCNN

作　　者：李东洁[1] 李若昊 Li Dongjie;Li Ruohao(School of Automation,Harbin University of Science and Technology,Harbin,Heilongjiang 150080,China)

机构地区：[1]哈尔滨理工大学自动化学院,黑龙江哈尔滨150080

出　　处：《激光与光电子学进展》2020年第4期345-352,共8页Laser & Optoelectronics Progress

基　　金：国家自然科学基金青年基金(51105117);黑龙江省普通高校基本科研业务费专项基金(LGYC2018JQ017)。

摘　　要：Faster RCNN在小目标的检测方面准确度和鲁棒性性能不够理想。为此,提出一种改进Faster RCNN对马克杯缺陷进行检测,将Faster RCNN和特征金字塔网络(FPN)相结合,增加使用细节化的浅层特征,使其对小目标有更好的检测效果。分别使用改进前后的Faster RCNN在Caffe上进行仿真。结果表明,Faster RCNN检测马克杯缺陷的性能出色,但是遗漏了部分小目标;改进后的Faster RCNN对缺口和划痕缺陷的检测准确率最多提升2.485个百分点,并且在小目标识别方面也有更优异的表现。Faster RCNN has poorer performance in terms of accuracy and robustness for detecting small targets.For this reason,an improved Faster RCNN was proposed to detect the defects in mugs.The Faster RCNN and feature pyramid network(FPN)were combined to increase the use of detailed shallow features,so as to achieve better detection effect for small targets.Faster RCNNs before and after improvement were used to conduct simulation on Caffe.The results show that Faster RCNN performs well in defect detection for mugs,but it misses some small targets.The improved Faster RCNN increases the detection accuracy by 2.485percent at most for gaps and scratches and performs better in small target recognition.

关 键 词：机器视觉 深度学习 FASTER RCNN 特征金字塔网络 缺陷检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]