第14卷第2期 智能系统学报 Vol.14 No.2 2019年3月 CAAI Transactions on Intelligent Systems Mar.2019 D0:10.11992/tis.201710005 网络出版地址：http:/kns.cnki.net/kcms/detail/23.1538.TP.20180413.0946.004html 卷积神经网络的贴片电阻识别应用 谌贵辉，何龙，李忠兵，亢宇欣，江枭宇 (西南石油大学电气信息学院，四川成都610500) 摘要：贴片电阻由于其体积微小、性能稳定等独特的性质，在当今智能化的电子设备中被广泛使用。为保证 贴片电阻的出厂质量，需要对其进行缺陷识别、极性方向识别、正反面识别和种类识别，目前很大程度上依靠 人工肉眼进行识别检测，效率低、容易误检、成本高。本文针对传统图像识别方法的局限性，结合近年来卷积 神经网络在图像识别方面所取得的巨大成就，基于AlexNet模型、GoogLeNet模型、ResNet模型思想设计了3种 深度适宜、可训练参数约4×10（百万）的卷积神经网络，克服了当前主流卷积神经网络模型由于可训练参数过 多、模型层数太深导致在贴片电阻识别应用中识别速度不能满足实时性要求、泛化识别准确率低的问题。实 验表明，3种模型的识别准确率均超过90%.最高识别准确率达到95%.识别速度达到0.203s/张(256像 素×256像素，COE5)。因此，本文设计的3种卷积神经网络可根据具体实际需求进行选用，在实践中具有极 强的可行性和可推广性，同时也在提升企业生产效率和产品质量方面具有重要意义。 关键词：贴片电阻识别；卷积神经网络；AlexNet模型；GoogLeNet模型；ResNet模型 中图分类号：TP391文献标志码：A文章编号：1673-4785(2019)02-0263-10 中文引用格式：谌贵辉，何龙，李忠兵，等.卷积神经网络的贴片电阻识别应用.智能系统学报，2019,142)：263-272. 英文引用格式：CHEN Guihui,HE Long,.LI Zhongbing,etal.Chip resistance recognition based on convolution neural network J, CAAI transactions on intelligent systems,2019,14(2):263-272. Chip resistance recognition based on convolution neural network CHEN Guihui,HE Long,LI Zhongbing,KANG Yuxin,JIANG Xiaoyu (School of Electrical Information,Southwest Petroleum University,Chengdu 610500,China) Abstract:Chip resistors are widely used in intelligent electronic devices because of their unique properties such as small size and stable performance.The chip resistors produced by the factory must be identified for defects in both front and back faces,polarity,and type in order to guarantee the quality.However,such identification largely relies on the eye de- tection,which is inefficient,prone to error,and costly.In this paper,considering the limitation of the traditional image recognition methods and the great achievements of convolutional neural network(CNN)in image recognition in recent years,three CNN models,AlexNet model,GoogLeNet model,and ResNet model,with appropriate depth and training parameters of about 4M(million)are designed to overcome the demerits of low speed that results in the inability to meet the real-time requirement.These models overcome the low accuracy problem of generalization recognition associated with the prevailing CNN models,which is caused by many trainable parameters and many layers of model.Experiments show that the recognition accuracy of these three models exceeds 90%.The highest recognition accuracy rate is 95%, and the recognition speed is 0.203 s/piece(256 x 256 pixels,CORE I5).Therefore,these three CNN models can be ad- opted in practice and have a strong feasibility and replicability;thus,they have a great potential to improve the produc- tion efficiency and product quality for chip resistors. Keywords:Chip resistance recognition:convolution neural network:AlexNet model:GoogLeNet model:ResNet model 收稿日期：2017-10-11.网络出版日期：2018-04-13. 基金项目：四川省科技支撑计划项目(2016GZ0107);四川省教 当今正处于信息智能时代，电子元器件犹如 育厅重点项目(16ZA0065):南充市重点科技项目 构筑起这个时代的一块块砖瓦，唯有对这些砖瓦 (NC17SY4001). 通信作者：何龙.E-mail:396024902@q9.com. 的质量进行保证，才能坚固地铸就属于这个时代DOI: 10.11992/tis.201710005 网络出版地址: http://kns.cnki.net/kcms/detail/23.1538.TP.20180413.0946.004.html 卷积神经网络的贴片电阻识别应用 谌贵辉，何龙，李忠兵，亢宇欣，江枭宇 （西南石油大学 电气信息学院，四川 成都 610500） 摘 要：贴片电阻由于其体积微小、性能稳定等独特的性质，在当今智能化的电子设备中被广泛使用。为保证 贴片电阻的出厂质量，需要对其进行缺陷识别、极性方向识别、正反面识别和种类识别，目前很大程度上依靠 人工肉眼进行识别检测，效率低、容易误检、成本高。本文针对传统图像识别方法的局限性，结合近年来卷积 神经网络在图像识别方面所取得的巨大成就，基于 AlexNet 模型、GoogLeNet 模型、ResNet 模型思想设计了 3 种 深度适宜、可训练参数约 4×106 (百万) 的卷积神经网络，克服了当前主流卷积神经网络模型由于可训练参数过 多、模型层数太深导致在贴片电阻识别应用中识别速度不能满足实时性要求、泛化识别准确率低的问题。实 验表明，3 种模型的识别准确率均超过 90%，最高识别准确率达到 95%，识别速度达到 0.203 s/张 (256 像 素×256 像素，CORE I5)。因此，本文设计的 3 种卷积神经网络可根据具体实际需求进行选用，在实践中具有极 强的可行性和可推广性，同时也在提升企业生产效率和产品质量方面具有重要意义。 关键词：贴片电阻识别；卷积神经网络；AlexNet 模型；GoogLeNet 模型；ResNet 模型 中图分类号：TP391 文献标志码：A 文章编号：1673−4785(2019)02−0263−10 中文引用格式：谌贵辉, 何龙, 李忠兵, 等. 卷积神经网络的贴片电阻识别应用[J]. 智能系统学报, 2019, 14(2): 263–272. 英文引用格式：CHEN Guihui, HE Long, LI Zhongbing, et al. Chip resistance recognition based on convolution neural network[J]. CAAI transactions on intelligent systems, 2019, 14(2): 263–272. Chip resistance recognition based on convolution neural network CHEN Guihui，HE Long，LI Zhongbing，KANG Yuxin，JIANG Xiaoyu (School of Electrical Information, Southwest Petroleum University, Chengdu 610500, China) Abstract: Chip resistors are widely used in intelligent electronic devices because of their unique properties such as small size and stable performance. The chip resistors produced by the factory must be identified for defects in both front and back faces, polarity, and type in order to guarantee the quality. However, such identification largely relies on the eye de￾tection, which is inefficient, prone to error, and costly. In this paper, considering the limitation of the traditional image recognition methods and the great achievements of convolutional neural network (CNN) in image recognition in recent years, three CNN models, AlexNet model, GoogLeNet model, and ResNet model, with appropriate depth and training parameters of about 4M (million) are designed to overcome the demerits of low speed that results in the inability to meet the real-time requirement. These models overcome the low accuracy problem of generalization recognition associated with the prevailing CNN models, which is caused by many trainable parameters and many layers of model. Experiments show that the recognition accuracy of these three models exceeds 90%. The highest recognition accuracy rate is 95%, and the recognition speed is 0.203 s/piece (256 × 256 pixels, CORE I5). Therefore, these three CNN models can be ad￾opted in practice and have a strong feasibility and replicability; thus, they have a great potential to improve the produc￾tion efficiency and product quality for chip resistors. Keywords: Chip resistance recognition; convolution neural network; AlexNet model; GoogLeNet model; ResNet model 当今正处于信息智能时代，电子元器件犹如 构筑起这个时代的一块块砖瓦，唯有对这些砖瓦 的质量进行保证，才能坚固地铸就属于这个时代 收稿日期：2017−10−11. 网络出版日期：2018−04−13. 基金项目：四川省科技支撑计划项目 (2016GZ0107)；四川省教 育厅重点项目 (16ZA0065)；南充市重点科技项目 (NC17SY4001). 通信作者：何龙. E-mail：396024902@qq.com. 第 14 卷第 2 期 智 能 系 统 学 报 Vol.14 No.2 2019 年 3 月 CAAI Transactions on Intelligent Systems Mar. 2019