Material classification in X-ray images based on multi-scale CNN

Abstract Security X-ray baggage scanners provide images based on the different levels of radiation absorption by materials. Images captured such are inspected a human operator, which can slow down verification process. To speed up inspection time, computer vision and machine learning methods increasingly being used. While object recognition has been subject huge number articles, problem material in still requires some work to achieve equivalent accuracy. This paper focuses discrimination materials into several classes, as organic substances or metals, obtained from dual-energy security scanners. We propose new multi-scale convolutional neural network (CNN) for predicting class, five sizes patches implemented parallelly balance trade-off between increase receptive field loss detail. analyze regularization activation functions their impact effectiveness our architecture. The results were compared with other popular CNN architectures demonstrate superiority solution.