Research on Double-layer Feature Classification Algorithm for Liquid Dangerous Goods Detection Based on UWB Centimeter Wave

In recent years, with the rapid development of the economy and the frequent flow of logistics, the common security inspection technology has low efficiency and unstable accuracy in the detection of liquid dangerous goods. This study proposes a two-layer feature extraction algorithm based on UWB centimeter-wave detection, which is composed of shallow autoencoder and deep LSTM network. In order to abstract the best description feature, the shallow autoencoder adds a classification constraint. In the classification stage, the deep algorithm LSTM can further abstract the sequence composed of shallow features into deep features to improve the accuracy of classification. The experimental results show that the autoencoder with shallow constraints is more suitable for feature extraction of UWB centimeter-wave signals in this experimental scene than PCA and ICA feature extraction algorithms. Compared with KNN, linear kernel SVM, Gaussian kernel SVM, decision tree and AdaBoost algorithms for sequence processing, LSTM has better processing effect and higher accuracy of classification accuracy. By comparing the test accuracy of other algorithms, the double-layer feature classification algorithm performs better in this experimental scene. The final test accuracy can reach more than 90%.