基于神经网络的塑料打包带高光谱模式识别

目的 建立一种快速、准确、无损的塑料打包带的检验及分类方法。方法 利用高光谱在波长为350～990 nm的条件下采集52个不同来源的塑料打包带样品的高光谱数据，并对样品进行Savitzky-Golay平滑处理，同时结合主成分分析对样品进行降维。将提取到的主成分进行K-Means聚类，以聚类结果为依据建立径向基函数神经网络（RBFNN）与BP神经网络模型（BPNN）。结果 打包带样品的高光谱谱图在400～500 nm、600～700 nm处有较大区别。实验共提取了5个初始特征值大于1的主成分，可以解释96.633%的原始数据。通过K-means聚类将塑料打包带样品分为6类，Calinski-Harabasz指数为28.76,RBFNN分类准确率为86.7%;BPNN分类准确率为98.1%,BPNN的分类效果更好。结论 研究表明神经网络在高光谱谱图分类处理上具有较高的准确度，同时也验证了高光谱在区分检验塑料打包带类物证的可行性与科学性，为公安机关提供了一种新的检验方法。

Hyperspectral Pattern Recognition of Plastic Packaging Tape Based on Neural Network

The work aims to establish a fast, accurate, and non-destructive inspection and classification method for plastic packaging tapes. 52 samples of plastic packaging tape were collected from different sources through hyperspectral data in the wavelength range of 350-990 nm, and the samples were smoothed with Savitzky Golay. Principal component analysis was also used to reduce the dimensionality of the samples. K-Means clustering was conducted on the extracted principal components, and a radial basis function neural network (RBFNN) and BP neural network model (BPNN) was established based on the clustering results. There were significant differences in the hyperspectral spectra of the packaged sample at 400-500 nm and 600-700 nm. A total of 5 principal components with initial feature values greater than 1 were extracted in the experiment, which could explain 96.633% of the original data. The plastic packaging tape samples were clustered into 6 categories, with a Calinski Harabasz index of 28.76 for K-means and a classification accuracy of 86.7% for RBFNN. The classification accuracy of BPNN was 98.1%. BPNN had better classification performance. Research has shown that neural network has high accuracy in the classification and processing of hyperspectral spectra, and it has also verified the feasibility and scientificity of hyperspectral recognition in the detection of plastic packaging tape type evidence, providing a new inspection method for public security organs.