The Recognition Method of Express Logistics Restricted Goods Based on Deep Convolution Neural Network

摘要

With the continuous increase of China’s express logistics business volume and the emergence of various security risks, social concern to the issue of express safety has been raised. Based on deep convolutional neural network, this study introduced an automatic identification method for express logistics restriction products. This method establishes a deep neural network structure model including convolutional blocks and fully connected layers by using the express package X-ray image dataset. It achieves 93.1% recognition accuracy, and 853ms of single image recognition time, which is far superior to the traditional manual detection speed. Compared with other traditional identification methods, for example, random forest, decision tree, Bayesian network and so on, this method not only guarantees real-time identification, but also significantly improves the recognition accuracy by about 20%.