In order to establish a fast and non-destructive analytical method for plastic bottle cap inspection,a hyper-spectral imaging system was used to inspect 48 plastic bottle cap samples.Firstly,the original spectra were prepro-cessed,and then principal component analysis,partial least squares-discriminant analysis,and competitive adaptive re-weighted sampling were used to construct hyperspectral datasets.Support vector machines,multi-layer perceptron mod-els,and convolutional neural networks were used to train the datasets.The results show that the hyperspectral images of plastic bottle caps constructed using competitive adaptive reweighting sampling extraction achieved an accuracy of 100％in the test set of convolutional neural networks.This method is convenient,fast,non-destructive,and requires minimal usage,providing strong support for the classification of plastic bottle caps.
维普
万方数据