基于分析谱图识别比对的高分子材料一致性判定系统的研究

发布时间：2018-04-04 20:51

本文选题：图像识别　切入点：一致性判定　出处：《北京化工大学》2016年硕士论文

【摘要】：高分子材料因其来源可靠、性质多样而在各类产品制造中获得了极其广泛的应用,其品质一致性监控也因此成为产品质量监控的一个重要环节。一直以来,质监部门的专家通过人工比对标准样品与送检样品的红外(IR)、差示(DSC)和热重(TGA)三类分析谱图,判定在用高分子材料与报批高分子材料的一致性,工作量巨大,人工成本高。本文以建立高分子材料一致性自动判定系统为目的,开展了以下研究：1.构造了分析谱图曲线的识别方法,该方法包括两个部分：(1)基于像素积分投影,确定坐标轴、原点、刻度位置并利用本文构造的特征数值识别方法,识别刻度数值,建立比例尺——被测物理量和像素间的换算关系。(2)利用颜色特征提取曲线像素点,经比例尺换算得到曲线的物理数值。2.构造了分析谱图特征数值的识别方法,该方法包括两个部分：(1)对相对单纯的IR谱图,采用像素积分投影法快速定位和截取特征数值图像；对背景复杂的DSC、TGA谱图,采用连通域法首先定位特征数值的单位符号,再根据符号与数值的位置关系,定位和截取谱图中的特征数值图像。(2)将截取的特征数值图像切分成多个单字符,提取各字符的均匀网格特征,通过基于距离分类的模板匹配辨识字符,将辨识得到的字符依顺序组合得到特征数值。3.基于识别出的曲线和特征数值,运用高、低阈值设置,建立了IR、DSC以及TGA三类谱图的一致性判定方法。对三类共计1318组谱图进行测试,结果表明,该方法发生错误判定的可能性极低。4.基于三类分析谱图的一致性判定结果,建立了高分子材料样品一致性判定方法。该方法可对91%以上的样品得出确定性判定结果,确定性判定结果的准确率达99%以上。5.建立了高分子材料一致性自动判定的应用系统。该系统基于B/S架构,提供了“用户管理”,“默认参数设置”,“一致性判定”等功能,投用以来运行稳定,取得了良好的应用效果。
[Abstract]:Polymer materials have been widely used in the manufacture of various products because of their reliable sources and diverse properties. Therefore, the quality consistency monitoring has become an important part of product quality monitoring.All along, experts in quality supervision departments have judged the consistency between the use of polymer materials and the approval of polymer materials through manual comparison of three kinds of analytical spectra of standard samples and samples submitted for inspection by infrared IR, differential scanning spectroscopy (DSCS) and thermogravimetric analysis (TGA), the workload of which is enormous.The cost of labor is high.In order to establish an automatic system for determining the consistency of polymer materials, the following research work has been carried out in this paper: 1: 1.A recognition method for analyzing spectral curve is constructed. The method includes two parts: 1) based on pixel integral projection, the coordinate axis, origin, calibration position are determined and the scale value is recognized by using the characteristic numerical recognition method constructed in this paper.The conversion relation between the measured physical quantity and the pixel is established. (2) the curve pixel is extracted by using the color feature, and the physical value of the curve is obtained by the scale conversion.In this paper, a method for identifying characteristic values of spectral maps is constructed. The method consists of two parts: 1) for relatively simple IR spectra, the pixel integral projection method is used to quickly locate and intercept the characteristic numerical images, and for the DSC-TGA spectra with complex background,The connected domain method is used to locate the unit symbols of the eigenvalues first, and then, according to the position relationship between the symbols and the values, the feature numerical images in the spectral map are located and intercepted.The uniform mesh feature of each character is extracted, and the recognition character is matched by template based on distance classification, and the character value. 3 is obtained by combining the identified characters in sequence.Based on the identified curves and characteristic values, a method for determining the consistency of three kinds of spectral diagrams, I. e., TGA and TGA, is established by setting up high and low threshold values.A total of 1318 sets of spectral diagrams were tested. The results show that the probability of error determination in this method is very low. 4. 4.Based on the results of consistency determination of three kinds of analytical spectra, a method for determining the consistency of polymer samples was established.This method can obtain deterministic results for more than 91% samples, and the accuracy of deterministic determination is over 99%.An application system for automatic determination of polymer material consistency is established.Based on the B / S architecture, the system provides functions such as "user management", "default parameter setting", "consistency judgment" and so on.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TB324;TP391.41

【相似文献】