Classification of X-ray scan images containing usb memory byartificial neural network-based image classification technique

As a result of the development and widespread use of X-ray scanning technology, the number of images produced is increasing at the same rate. X-ray scanning systems are used in many fields from medicine to industry and are frequently preferred in the field of security. New generation X-ray devices, which have the capacity to produce high resolution images with the developing technology, are of vital importance in the detection of potential threats in the field of security. However, the images produced by these devices, which can scan quickly during use, need to be quickly analyzed by the users. The rapid detection of small devices such as USB memory sticks, which threaten data security, from X-ray images poses a problem. This study aims to detect whether there is USB memory in scan images with deep learning-based classification methods. In this context, data with or without 1217 USB sticks were created with X-Ray scanning systems. The created data set is trained with 8 different models with Convolutional Neural Networks (CNN) based deep learning architecture. As a result of the training, the ResNet50V2 model achieved the highest overall accuracy with a success rate (93%), while the lowest (67%) overall accuracy value was obtained with the ResNet50 model. Keywords: X-ray, Scanning Systems, Artificial Intelligence, Artificial Neural Networks, USB Memory, CNN