基于深度学习与工业服务器的云检测系统应用研究

Research on the Application of Cloud Detection System Based on Deep Learning and Industrial Server

为了解决工业质检场景中误检率高、直通率低,以及导入检测设备难升级的问题,基于局域网工业服务器的硬件方案,收集误检图像和缺陷图像,对其完成标注,并以负载均衡和并发稳定为目的规划云检测系统。基于深度学习的人工智能降误判模型,对各种误判图像进行学习训练,建立高直通率、低误判的检测机制。首先,进行以服务器与通信媒介为基础的云检测系统硬件选型,建立上下位机的数据通信规则、数据发送接收格式规则。然后,开发云检测系统的五大模块:上位机接收模块、上位机分析模块、上位机发送模块、下位机发送模块、下位机接收模块,整合为一个云检测系统平台。最后,基于大数据缺陷学习,调用AI学习软件平台生成的降误判模型,供云检测系统使用,并集成在系统中。实验测试结果显示:系统有利于在设备改造升级场景中降低缺陷误判,为高性能云检测系统的推广奠定解决方案基础。

To solve the problems of high false detection rate and low first pass yield rate in industrial quality inspection scenarios,as well as the difficulty of upgrading imported inspection equipment,the research collects false detection images and defect images based on the hardware solution of a local area network industrial server,labels them,and plans a cloud detection system for the purpose of load balancing and concurrency stability.The artificial intelligence model for reducing false positives based on deep learning conducts learning and training on various false positives,and establishes a detection mechanism with high first pass yield rate and low false detection rate.First,conduct hardware selection for cloud detection systems based on servers and communication media,and establish data communication rules and data transmission and reception format rules for upper and lower computers.Then,five modules of the cloud detection system are developed:the upper computer receiving module,the upper computer analysis module,the upper computer sending module,the lower computer sending module,and the lower computer receiving module,which are integrated into a cloud detection system platform.Finally,based on big data defect learning,the error reduction model generated by the AI learning software platform is invoked for use by the cloud detection system and integrated into the system.The experimental test results show that the system in this article is conducive to reducing misjudgment in equipment upgrading scenarios,laying a solution foundation for the promotion of high-performance cloud detection systems.