红外光谱结合化学计量学对无纺布袋的分类研究

Classification of Non-woven Fabric Bags by Infrared Spectroscopy Combined with Chemometrics

  • 摘要: 建立一种快速无损的对无纺布袋的分类方法.利用便携式傅里叶变换红外光谱法对收集到的 54 个不同品牌、不同规格、不同颜色的无纺布袋样品进行测试,并结合化学计量方法对样品进行分组.最后采用神经网络MLP模型和RBF模型对分类结果进行验证.依据 54 个无纺布袋样品的红外光谱图数据可以将样品分为三大类,选取第Ⅰ类样品采用系统聚类和皮尔逊相关性的距离矩阵方法可以分为 8 组.对无纺布袋样品的红外光谱数据分别进行 FDA处理和 PCA降维后建立神经网络模型,MLP模型对PCA降维后的数据分类准确率为 97%,RBF模型对PCA降维后的数据分类准确率为 93%,通过数据对比得到,MLP模型的分类结果更加准确.该方法简便易行,科学可靠,为公安基层工作提供了新方法和新思路,在实际公安工作中具有较好的应用价值.

     

    Abstract: To establish a fast and non-destructive classification method for non-woven fabric bags.The method used portable Fourier transform infrared spectroscopy to test 54 samples of non-woven bags of different brands,specifications,and colors collected.The samples were preliminarily classified based on infrared spectra and grouped using stoichiometric methods.Finally,the classification results were validated using the neural network MLP model and RBF model.Results According to the infrared spectrum data of 54 non-woven bag samples,the samples can be divided into three categories,and the first category samples can be divided into eight groups by using hierarchical clustering and Pearson correlation Distance matrix method.After FDA processing and PCA dimensionality reduction,a neural network model was established for the infrared spectral data of non-woven bag samples.The MLP model had a classification accuracy of 97%for PCA dimensionality reduction data,while the RBF model had a classification accuracy of 93%for PCA dimensionality reduction data.Through data comparison,it was found that the MLP model had a more accurate classification result.This method is simple,feasible,scientific,and reliable,providing new methods and ideas for grassroots public security work,and has good application value in practical public security work.

     

/

返回文章
返回