摘要
目前超市收银系统大多采用收银员手持扫码枪逐一识别商品条形码的传统方式,效率相对较低。针对此现状,笔者基于深度学习中的图像识别SSD算法,以Xilinx公司的PYNQ-Z2为开发平台,设计了一款智能超市收银系统。本系统由商品信息采集模块,商品检测模块和上位机定价模块三部分组成。商品信息采集模块通过摄像头采集检测区域内商品的图像,通过USB接口传输到PYNQ的ARM端;商品检验模块负责在PYNQ平台上预处理图像大小和格式调整。处理后传输给上位机,上位机将预处理后的商品图像与已训练好的商品模型进行匹配识别,并将检测结果反馈给计价模块;计价模块一方面展示识别出的商品图像以及商品相关信息,一方面实现结算、清零等功能。本系统能一次检测多个商品并同步结算,为超市收银方式提供了一种新方案。
At present, most supermarket cash register systems use the traditional way of cashiers' hand-held scanning code guns to identify commodity barcodes one by one, and the efficiency is relatively low. Aiming at this situation, based on the image recognition SSD algorithm in deep learning, and with Xilinx PYNQ-Z2 as the development platform, a smart supermarket cash register system is designed. The system consists of three parts including commodity information collection module, commodity detection module and upper computer pricing module. The commodity information collection module collects the image of the commodity in the detection area via the camera, and transmits it to the ARM end of the PYNQ via the USB interface;The commodity inspection module is responsible for pre-processing image size and format adjustments on the PYNQ platform, and after processing, the information is transmitted to the upper computer, and the upper computer matches and identifies the pre-processed product image with the trained commodity model, and feeds the detection result to the pricing module;The pricing module performs functions such as settlement and clearing while displaying the recognized product image and product related information. This system can detect multiple commodities at one time and settle accounts synchronously, providing a new scheme for supermarket cash-collecting mode.
作者
孙璐阳
SUN Lu-yang(Shandong University of Science and Technology,Qingdao Shandong 266590,China)
出处
《通信技术》
2019年第5期1279-1283,共5页
Communications Technology
