基于EasyDL提升回收卷烟包装箱效率的功能实现

Research on Improving the Efficiency of Recycled Cigarette Packaging Box Based on EasyDL

目的:通过自动化机器学习技术,提升回收卷烟包装箱分拣工序的自动化程度和准确率。方法基于百度EasyDL定制化图像识别框架,构建基于机器视觉设计的深度学习分拣控制系统。该功能模块主要包括PC一体机、光感应模块、工业相机、条码识别模块、分拣控制模块、声光报警模块等。重点探讨了分拣过程中,包装箱未贴标签,包装箱混装,标签位置不标准等问题的识别判断。结果:依据针对目前使用的图片数据集及质量的高低,使用了EasyDL定制化图像识别框架所构建的深度学习模型流水作业系统,分类准确率可以达到99%以上。结论:基于EasyDL机器视觉学习卷烟包装箱回收分拣模块,能够满足分拣要求,替代人工完成自动化分拣,简化深度学习门槛,降低企业成本。

To improve the automation and accuracy of the sorting process of recycled cigarette packaging boxes through automated machine learning technology.Based on Baidu EasyDL customized image recognition framework,a deep learning sorting control system based on machine vision design was constructed.This functional module mainly includes PC integrated machine,light sensor module,industrial camera,barcode recognition module,sorting control module,sound and light alarm module,etc.The focus is on the identification and judgment of unlabeled packaging boxes,mixed packaging of packaging boxes,and non-standard label positions during the sorting process.According to the currently used image data set and quality,the deep learning model pipeline system constructed by the EasyDL customized image recognition framework was used,and the classification accuracy rate could reach more than 99%.Based on the EasyDL machine vision learning cigarette packaging box recycling and sorting module,it can meet the sorting requirements,replace manual automatic sorting,simplify the deep learning threshold,and reduce enterprise costs.