Improving Content Recognition of X-ray Images of Baggage Using Deep CNN in Customs Administration

Abstract

With the increasing security threats and the need for ensuring the safety of the movement of passengers and goods through countries, customs inspections at airports and border crossings are more important than ever. Customs departments currently rely mainly on visual inspection to detect dangerous and prohibited items in X-ray images of baggage. However, this traditional approach faces many challenges such as the possibility of human errors and the long time spent in the inspection process, in addition to the possibility of human errors in analyzing and interpreting complex X-ray images. Therefore, the need has emerged to use advanced artificial intelligence techniques to improve the speed and accuracy of detecting prohibited materials. This research seeks to improve the recognition of the content of X-ray images of baggage in customs administration using deep convolutional neural network models of artificial intelligence. A number of experiments have been carried out to improve the accuracy and detection of dangerous materials, especially firearms, sharp materials and knives, through automated X-ray images of baggage using three models of convolutional neural networks which are VGG16, ResNet50, and InceptionV3. The aim of using different models is to obtain the highest accuracy among them. The three models have been trained and tested using a huge SIXray dataset, which specializes in X-ray images of baggage. The results show that the VGG16 model has outperformed the others with a high accuracy exceeding 96%. The contribution of this research is to enhance the efficiency of customs inspection operations and improving security and safety levels through accurate and rapid automated detection of dangers materials to prevent them from being entered to the country.