基于深度学习的商品识别方法研究
|
摘要
无人零售行业异常火爆,对零售商品的识别是技术关键。便利店场景下实现无人零售有很大的需求。传统方式是条码配合扫码器,不过用户体验不太好;无线射频识别技术(RFID)的标签成本和贴标成本较高,所以尝试计算机视觉和深度学习结合的商品识别方法。运用深度学习中的AlexNet神经网络模型,通过商品图像预处理以及图像旋转、镜像翻转等方式扩充数据集,再利用随机梯度下降法(SGD)进行模型算法优化,生成对50种常用商品图像的识别模型。结果表明,利用深度学习对商品图像进行分类识别有较好的结果。
Unattended retail industry is extremely hot,the identification of retail products is the key technology. Convenience store scene to achieve unattended retail has great demand. The traditional method is bar code with the codec,but the user experience is not very good and learning is more difficult; RFID tags have high labeling costs and labeling costs,so try combining computer vision with deep learning Product identification method. Using the AlexNet neural network model in depth learning, the data set was augmented by commodity image preprocessing, image rotation and mirroring, and then the algorithm was optimized by stochastic gradient descent(SGD) to generate 50 common commodity image recognition models. The results show that the use of depth learning product image classification has good results.
Unattended retail industry is extremely hot,the identification of retail products is the key technology. Convenience store scene to achieve unattended retail has great demand. The traditional method is bar code with the codec,but the user experience is not very good and learning is more difficult; RFID tags have high labeling costs and labeling costs,so try combining computer vision with deep learning Product identification method. Using the AlexNet neural network model in depth learning, the data set was augmented by commodity image preprocessing, image rotation and mirroring, and then the algorithm was optimized by stochastic gradient descent(SGD) to generate 50 common commodity image recognition models. The results show that the use of depth learning product image classification has good results.
引文
[1]张晓惠,郑敏东,刑志国.新型商品识别技术——射频识别技术[J].广东印刷,2005(04):54-55.
[2]郑建彬,白雅贤,詹恩奇,等.基于改进SIFT匹配方法的货架乳制品识别[J].计算机科学,2017,44(09):315-319.
[3]陈哲凡,郑建彬,詹恩奇,等.基于聚类SURF特征的商品识别算法[J].计算机应用研究,2017,34(11):3490-3493.
[4]曾志,吴财贵,唐权华,等.基于多特征融合和深度学习的商品图像分类[J].计算机工程与设计,2017,38(11):3093-3098.
[5]程欣.基于深度学习的图像目标定位识别研究[D].成都:电子科技大学,2016.
[6] LeCun Y,Bengio Y,Hinton G. Deep learning[J]. Nature,2015,521(7553):436-44.
[7]周志华.机器学习[M].北京:清华大学出版社,2016:97-118.
[8]代贺,陈洪密,李志申.基于卷积神经网络的数字识别[J].贵州师范大学学报:自然科学版,2017,35(05):96-101.
[9]杨帅,张有芬,李玉惠,等.基于深度卷积神经网络的车标分类[J].工业仪表与自动化装置,2017(05):75-78.
[10] Krizhevsky A,Sutskever I,Hinton G E. ImageNet clas-sification with deep convolutional neural networks[C].//International Conference on Neural InformationProcessing Systems. Curran Associates Inc. 2012:1097-1105.
[11]袁银,王东斌,刘永金.基于深度学习的植物图像识别方法研究[J].现代农业科技, 2017(23):278-280.
[12]吴超,邵曦.基于深度学习的指静脉识别研究[J].计算机技术与发展,2018(02):1-6.
[13] Smirnov E A,Timoshenko D M,Andrianov S N. Com-parison of Regularization Methods for ImageNet Classifi-cation with Deep Convolutional Neural Networks[J].Aasri Procedia,2014,6(1):89-94.
[14]梅啟成,王中任,何鑫,等.基于多幅标定板图像的摄像机标定的对比和分析[J].工具技术,2015,49(07):90-94.
[2]郑建彬,白雅贤,詹恩奇,等.基于改进SIFT匹配方法的货架乳制品识别[J].计算机科学,2017,44(09):315-319.
[3]陈哲凡,郑建彬,詹恩奇,等.基于聚类SURF特征的商品识别算法[J].计算机应用研究,2017,34(11):3490-3493.
[4]曾志,吴财贵,唐权华,等.基于多特征融合和深度学习的商品图像分类[J].计算机工程与设计,2017,38(11):3093-3098.
[5]程欣.基于深度学习的图像目标定位识别研究[D].成都:电子科技大学,2016.
[6] LeCun Y,Bengio Y,Hinton G. Deep learning[J]. Nature,2015,521(7553):436-44.
[7]周志华.机器学习[M].北京:清华大学出版社,2016:97-118.
[8]代贺,陈洪密,李志申.基于卷积神经网络的数字识别[J].贵州师范大学学报:自然科学版,2017,35(05):96-101.
[9]杨帅,张有芬,李玉惠,等.基于深度卷积神经网络的车标分类[J].工业仪表与自动化装置,2017(05):75-78.
[10] Krizhevsky A,Sutskever I,Hinton G E. ImageNet clas-sification with deep convolutional neural networks[C].//International Conference on Neural InformationProcessing Systems. Curran Associates Inc. 2012:1097-1105.
[11]袁银,王东斌,刘永金.基于深度学习的植物图像识别方法研究[J].现代农业科技, 2017(23):278-280.
[12]吴超,邵曦.基于深度学习的指静脉识别研究[J].计算机技术与发展,2018(02):1-6.
[13] Smirnov E A,Timoshenko D M,Andrianov S N. Com-parison of Regularization Methods for ImageNet Classifi-cation with Deep Convolutional Neural Networks[J].Aasri Procedia,2014,6(1):89-94.
[14]梅啟成,王中任,何鑫,等.基于多幅标定板图像的摄像机标定的对比和分析[J].工具技术,2015,49(07):90-94.