关键词: X射线荧光光谱, 人工神经网络, 判别分析, 系统聚类, 皮鞋

Abstract: In order to establish a classification and prediction model for quickly classifying and identifying the physical evidence of leather shoes and soles at the scene of the case, X-ray fluorescence spectroscopy is used and the working mode includes voltage 50 kV, current 200 μA. test time 60 s.The samples were analyzed and the components and contents of inorganic elements in 55 leather shoe sole samples were extracted. According to the types and contents of elements in the extracted samples, systematic clustering was carried out. Taking the results of systematic clustering as an example, three classification and prediction models of discriminant analysis, RBF neural network and MLP neural network were established. Finally, 55 samples were divided into 5 categories. After verification, the classification accuracy of multilayer perceptron neural network is relatively high, which is 94.4%. The results show that this method is fast and easy to operate, and the established classification prediction model can provide new ideas and methods for the actual handling of public security cases.

Key words: X-ray fluorescence spectrum, artificial neural network, discriminant analysis, systematic clustering, leather shoes