基于大数据驱动的电子商务产品自动化识别系统  

Automatic identification system of E⁃commerce products based on big data drive

作　　者：孟娟娜 MENG Juanna(University of Electronic Science and Technology of China,Chengdu 610054,China;Xi’an Vocational and Technical College,Xi’an 710077,China)

机构地区：[1]电子科技大学,四川成都610054 [2]西安职业技术学院,陕西西安710077

出　　处：《电子设计工程》2024年第20期89-92,共4页Electronic Design Engineering

摘　　要：受到摆放角度以及拍照角度等多种因素的影响,电子商务产品图像存在一定的角度偏差,为此设计基于大数据驱动的电子商务产品自动化识别系统。采用高清摄像头对产品图像进行采集,并构建设备层、传输层等硬件结构。构建灰度图像的分离指标,对产品图像进行二值化处理。计算产品图像的主方向角度,并结合角度矫正阈值,调整原始图像的角度。采用余弦距离作为特征向量相似度的判断标准,实现待识别图像与图像库的特征匹配。结合卷积网络,构建出产品分类识别模型,并对损失函数进行设计,从而实现产品自动化识别。测试结果表明,采用提出的系统识别结果的误匹配率较低,具备更高的识别精度。Influenced by many factors,such as placement angle and camera angle,the image of E-commerce products has certain angle deviation.Therefore,an automatic identification system of E-commerce products based on big data drive is designed.High-definition camera is used to collect product images,and hardware structures such as device layer and transport layer are constructed.The separation index of gray image is constructed,and the product image is binarized.Calculate the main direction angle of the product image,and adjust the angle of the original image in combination with the angle correction threshold.Cosine distance is used as the criterion to judge the similarity of feature vectors,and the feature matching between the image to be recognized and the image database is realized.Combined with convolution network,the product classification and identification model is constructed,and the loss function is designed,so as to realize automatic product identification.The test results show that the false matching rate of the recognition results with the proposed system is low and the recognition accuracy is higher.

关 键 词：大数据驱动 电子商务 产品识别 识别精度 

分 类 号：TN609[电子电信—电路与系统]