BOPET薄膜杂质在线检测仪的研究

发布时间：2018-04-03 02:31

本文选题：DM6446　切入点：薄膜杂质识别　出处：《哈尔滨理工大学》2015年硕士论文

【摘要】：随着图像采集技术与数字图像处理研究水平不断提高,薄膜杂质检测设备也在不断的更新,所以薄膜生产厂家对准确性高、可靠性好的BOPET薄膜杂质在线检测仪的需求量大而且要求也更高。本文采用了基于LVQ的薄膜杂质检测与识别算法,搭建了以TI公司TMS320DM6446处理器为核心的硬件平台的在线检测仪。本文通过对CCD采集到BOPET薄膜图像进行预处理获得杂质轮廓,提取特征值作为LVQ神经网络的输入向量,通过学习样本使得LVQ神经网络具备识别杂质的能力,从而可以得到薄膜杂质的信息。将此在线检测仪应用在薄膜生产环节中,可以实现生产管理的自动化,提高生产效率。本文首先介绍了薄膜杂质检测的背景及研究现状,对目前的一些薄膜杂质检测技术及实现方法进行比较,再分析了本系统的设计需求,最终确定了基于LVQ神经网络的薄膜杂质检测与识别算法的在线检测系统方案。本系统的硬件采用了模块化设计方法,其中包括微处理器模块、图像采集模块、外扩存储模块、实时数据缓存模块、数据传输模块、杂质标记模块、电源模块以及时钟模块。在软件开发过程中,选择了ARM端的Linux操作系统及DSP端的DSP/BIOS为软件开发平台,采用LVQ神经网络算法实现薄膜杂质目标的检测与识别。本文也对系统的平台移植作了介绍,并在两个子系统端都设计了系统软件。本系统在算法设计时采用了Matlab的仿真实验,同时在硬件平台搭建完成后也实现了板级的实验认证。实验表明,本系能够实现对BOPET薄膜杂质的在线检测与识别,并且适合于工业现场生产线流水作业中。
[Abstract]:With the continuous improvement of image acquisition technology and digital image processing, the film impurity detection equipment is also constantly updated, so the film manufacturers have high accuracy.The reliability of BOPET thin film on-line detector is in great demand and even higher.In this paper, the LVQ based thin film impurity detection and recognition algorithm is adopted, and an on-line detector based on TI's TMS320DM6446 processor is built.In this paper, the impurity profile is obtained by preprocessing the BOPET film image collected by CCD, and the eigenvalue is extracted as the input vector of the LVQ neural network. The LVQ neural network has the ability to recognize impurities by learning the samples.Thus, the information of impurity in the film can be obtained.The application of this on-line detector in film production can realize the automation of production management and improve production efficiency.This paper first introduces the background and research status of the film impurity detection, compares the current technology and implementation methods of the membrane impurity detection, and then analyzes the design requirements of the system.Finally, an online detection system based on LVQ neural network is proposed.The hardware of this system adopts modularization design method, including microprocessor module, image acquisition module, expandable storage module, real-time data buffer module, data transmission module, impurity mark module, power supply module and clock module.In the process of software development, the Linux operating system of ARM and the DSP/BIOS of DSP are selected as the software development platform, and the detection and recognition of impurity targets in thin film are realized by LVQ neural network algorithm.This paper also introduces the platform transplantation of the system, and designs the system software at the two subsystems.In this system, the Matlab simulation experiment is used in the algorithm design, and the board level experiment authentication is also realized after the hardware platform is built.The experimental results show that the system can detect and identify impurities in BOPET films on line and is suitable for industrial production line.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TB383.2

【相似文献】