Hyperspectral imaging technology for identification of polymeric plastic automobile lampshade

Abstract

The fragments of automobile lampshade widely appear in the scene of traffic accidents, homicide and other important cases. In this study, a scientific and effective method for rapid and non– destructive examination of this evidence by combining hyperspectral imaging (HSI) technology and deep learning was developed.45 lampshade samples of different and models were collected from different Auto repair shops. The hyperspectral imaging technology was used to collect the hyperspectral images and the reflectance spectral data from each of them. These datas of automobile lampshade samples were analyzed by chemometric methods, including K-nearest neighbor (KNN), support vector machine (SVM), convolutional neural network (CNN) and migration learning inception-resnet-v4 network and the results were discussed. The results showed that the KNN,SVM,CNN and the migration learning network models realized 84.0%,93.8%, 95.5%and97.3% accuracy for the6 categories of the 45 car lampshades with a total of 675 data samples. Based on the stringency and accuracy of the classification requirements, the migration learning inception-resnet-v4 network model was finally identified as the best model for the classification and recognition of automotive lampshades. The combination of hyperspectral imaging technology and deep machine learning achieved the purpose of distinguishing car lampshade brands, provided a potentially simple, non– destructive, and rapid method for automobile lampshades detection and classification.