基于机器视觉的纸张表面缺陷智能化图像处理研究
Design and Research of Intelligent Image Processing and Recognition System for Paper Surface Defects
王小春 1张宏甫1
作者信息
- 1. 西安航空职业技术学院,陕西 西安,710089
- 折叠
摘要
随着造纸工业自动化技术的快速发展,传统依赖人工视觉的纸张缺陷检测方式已显不足,机器视觉检测技术因此应运而生.针对纸张表面缺陷检测问题,提出一套先进的处理方案,包括图像预处理(噪声滤波、筛选、分割等步骤)及基于计算机视觉的缺陷识别体系架构设计.系统设计细节覆盖了软硬件选择及智能化图像处理算法的实现,特别是在缺陷特征提取方面,采用了形状特征与不变矩特征相结合的方法.此外,引入了遗传算法优化的支持向量机模型以提高分类精确度.实验证明,经过参数优化的分类器在缺陷检测上能达到 96.83%的高准确率,证明了该方法的有效性与实用性.
Abstract
With the rapid development of automation technology in the paper industry,the traditional paper defect detection method relying on artificial vision has been insufficient,so the machine vision detection technology has emerged.Aiming at the problem of paper surface defect detection,a set of advanced processing scheme is proposed,including image preprocessing(noise filtering,screening,segmentation,etc.)and flaw recognition architecture design based on computer vision.The details of the system design cover the selection of hardware and software and the realization of intelligent image processing algorithm,especially in the aspect of defect feature extraction,the method of combining shape feature and invariant moment feature is adopted.In addition,a support vector machine model optimized by genetic algorithm is introduced to improve the classification accuracy.Experimental results show that the classifier with optimized parameters can achieve 96.83%high accuracy in defect detection,which proves the effectiveness and practicability of this method.
关键词
纸张/缺陷/图像处理/缺陷分类/识别系统Key words
paper/defect/image processing/defect classification/identification system引用本文复制引用
出版年
2024
NETL
NSTL
万方数据