A Data Augmentation Method for Prohibited Item X-Ray Pseudocolor Images in X-Ray Security Inspection Based on Wasserstein Generative Adversarial Network and Spatial-and-Channel Attention Block

Abstract

For public security and crime prevention, the detection of prohibited items in X-ray security inspection based on deep learning has attracted widespread attention. However, the pseudocolor image dataset is scarce due to security, which brings an enormous challenge to the detection of prohibited items in X-ray security inspection. In this paper, a data augmentation method for prohibited item X-ray pseudocolor images in X-ray security inspection is proposed. Firstly, we design a framework of our method to achieve the dataset augmentation using the datasets with and without prohibited items. Secondly, in the framework, we design a spatial-and-channel attention block and a new base block to compose our X-ray Wasserstein generative adversarial network model with gradient penalty. The model directly generates high-quality dual-energy X-ray data instead of pseudocolor images. Thirdly, we design a composite strategy to composite the generated and real dual-energy X-ray data with background data into a new X-ray pseudocolor image, which can simulate the real overlapping relationship among items. Finally, two object detection models with and without our data augmentation method are applied to verify the effectiveness of our method. The experimental results demonstrate that our method can achieve the data augmentation for prohibited item X-ray pseudocolor images in X-ray security inspection effectively.