基于CNN与机器视觉的纸张尘埃度测量系统设计与研究

Design and Research of Paper Dustiness Measurement System Based on CNN and Machine Vision

关键词:  纸张尘埃度  卷积神经网络(CNN)  机器视觉  图像处理

Key Words:Paper dust content  Convolutional Neural Network (CNN)  Machine vision  Image processing

摘要:本研究设计了基于CNN与机器视觉的纸张尘埃度测量系统，实现了纸类产品尘埃度的高效、精准检测。该系统基于模型训练和检验两个模块构建，使用高分辨率扫描仪获取尘埃数据集和纸张试样图片，保证了测量精度。介绍了CNN尘埃图像分类过程及网络架构搭建策略，使用不同优化算法训练分类模型。采用了对角线测量算法，制作了标准尘埃像素表，可依据像素查值表进行定级和分类统计，进而计算尘埃度。结果表明，该系统的精度优于尘埃度检验标准要求，分类准确度达到95.89%，能够实现多类纸品的全量程测量，单样本重复性测量误差为0，与人工检测相比效率有显著提升。

Abstract:This study designed a paper dustiness measurement system based on Convolutional Neural Network (CNN) and machine vision, achieving efficient and accurate detection of the dustiness of paper products. The system is constructed with two modules: model training and verification. It uses a high - resolution scanner to obtain dust datasets and paper sample images, ensuring measurement accuracy. The process of CNN dust image classification and the strategy for building the network architecture are introduced, and different optimization algorithms are used to train the classification model. A diagonal measurement algorithm is adopted to create a standard dust pixel table, based on which the dust can be graded and statistically classified, and then the dustiness can be calculated. The results show that the accuracy of the system exceeds the requirements of the dustiness inspection standard, with a classification accuracy of 95.89%. It can achieve full-range measurement of various types of paper products, the repeatability error is 0, and there is a significant improvement in efficiency compared with manual detection.