基于嵌入式AI的缺陷检测标定系统  

Defect Detection and Calibration System Based on Embedded AI

作　　者：王硕 李艳秋 郭锋 杜茜 米富豪 夏敏耀 Wang Shuo;Li Yanqiu;Guo Feng;Du Qian;Mi Fuhao;Xia Minyao(School of Information Science and Engineering,Linyi University,Linyi,China)

机构地区：[1]临沂大学信息科学与工程学院,山东临沂

出　　处：《科学技术创新》2024年第20期97-100,共4页Scientific and Technological Innovation

基　　金：2023年山东省大学生创新创业训练计划项目(S202310452175)

摘　　要：为解决人工及传统方法对实木板材表面缺陷识别误差大、漏检率高等问题,本文设计了一款基于嵌入式AI的缺陷检测标定系统,用机器视觉技术代替传统检测,能够准确识别并标记木板的缺陷部分。系统基于YOLOv5目标检测算法并引入SE注意力机制,采用流水线20275张木材表面缺陷图像作为数据集,将模型部署到嵌入式AI平台上,当检测到缺陷时将缺陷信息发送给机械臂,并将像素坐标转换到实际坐标,实现对缺陷位置的标定。In order to solve the problem of large error and high missing rate of surface defect identification by manual and traditional methods,this paper designs a defect detection and calibration system based on embedded AI,which uses machine vision technology to replace traditional detection,and can accurately identify and mark the defect part of the board.The system is based on YOLOv5 target detection algorithm and introduces SE attention mechanism,uses 20,275 wood surface defect images of the pipeline as the data set,and deploy-es the model to the embedded AI platform.When the defect is detected,the defect information is sent to the mechanical arm,and the pixel coordinates are converted to the actual coordinates to realize the calibration of the defect position.

关 键 词：木材缺陷 YOLO 缺陷检测 机械臂 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.4[自动化与计算机技术—控制科学与工程] S781.5[农业科学—木材科学与技术]