基于视觉技术和深度学习的纸张缺陷识别系统及软件设计
Paper Defect Recognition System and Software Design Based on Visual Technology and Deep Learning
马静 1张志军1
作者信息
- 1. 陕西国防工业职业技术学院,陕西 西安,710300
- 折叠
摘要
为解决人工视觉不能识别纸张小尺寸缺陷,无法保障缺陷识别精确度与稳定性,且难以适应大面积、高速生产要求等一系列问题,以机器视觉技术与深度学习算法为载体进行了纸张缺陷识别系统硬件与软件设计,并提出了基于BP神经网络的纸张缺陷识别算法.其中系统硬件由工业摄像机、荧光灯、FPGA、DSP等构成,系统软件以MFC框架与C++编程语言为辅助开发设计.纸张缺陷识别算法则经过纸张缺陷图像预处理、特征提取与选择、识别分类实现缺陷识别.此系统识别方法简单、识别精准且实时性强,可在很大程度上取代传统人工视觉识别方法.
Abstract
In order to solve a series of problems such as the inability of artificial vision to recognize small-sized defects in paper,the inability to ensure the accuracy and stability of defect recognition,and the difficulty in adapting to the requirements of large-scale and high-speed production,this paper uses machine vision technology and deep learning algorithms as carriers to design the hardware and software of a paper defect recognition system,and proposes a paper defect recognition algorithm based on BP neural network.The system hardware is composed of industrial cameras,fluorescent lamps,FPGA,DSP,etc,and the system software is developed and designed with MFC framework and C++programming language as auxiliary,and the paper defect recognition algorithm is achieved through paper defect image preprocessing,feature extraction and selection,and recognition classification to achieve defect recognition.This system has a simple,accurate,and real-time recognition method,which can largely replace traditional artificial vision recognition methods.
关键词
机器视觉/深度学习/纸张缺陷/识别系统/软件设计Key words
machine vision/deep learning/paper defects/identification system/software design引用本文复制引用
基金项目
陕西省"十四五"教育科学规划2023年度课题(SGH23Y3055)
陕西省职业技术教育学会2023年度课题(2023SZX043)
出版年
2024
NETL
NSTL
万方数据