“Unexpected Item in the Bagging Area”: Anomaly Detection in X-Ray Security Images

Abstract

The role of Anomaly Detection in X-ray security imaging, as a supplement to targeted threat detection, is described; and a taxonomy of anomalies types in this domain is presented. Algorithms are described for detecting appearance anomalies, of shape, texture and density; and semantic anomalies of object category presence. The anomalies are detected on the basis of representations extracted from a convolutional neural network pre-trained to identify object categories in photographs: from the final pooling layer for appearance anomalies, and from the logit layer for semantic anomalies. The distribution of representations in normal data are modelled using high-dimensional, fullcovariance, Gaussians; and anomalies are scored according to their likelihood relative to those models. The algorithms are tested on X-ray parcel images using stream-of-commerce data as the normal class, and parcels with firearms present as examples of anomalies to be detected. Despite the representations being learnt for photographic images, and the varied contents of stream-ofcommerce parcels; the system, trained on stream-of-commerce images only, is able to detect 90% of firearms as anomalies, while raising false alarms on 18% of stream-of-commerce.

Meta-Transfer Learning Driven Tensor-Shot Detector for the Autonomous Localization and Recognition of Concealed Baggage Threats

“…The supervised deep learning methods to recognize baggage threats are based on object detection [ 32 , 33 , 34 ], classification [ 35 , 36 ], and segmentation [ 11 , 37 ] schemes. The majority of these methods also utilize one-staged [ 38 ] and two-staged [ 9 ] detectors such as YOLO [ 39 ], YOLOv2 [ 40 ], RetinaNet [ 41 ], and Faster R-CNN [ 42 ].…”

“…Gaus et al [ 10 ] evaluated RetinaNet [ 41 ], Faster R-CNN [ 42 ] and Mask R-CNN [ 45 ] backboned through ResNet-18 [ 46 ], ResNet-50 [ 46 ], ResNet-101 [ 46 ], SqueezeNet [ 47 ], and VGG-16 [ 48 ] for screening baggage X-ray scans as benign or malignant [ 10 ]. Griffin et al [ 36 ] classified unexpected items within the bagging areas based upon their shape, texture, and density, and semantic appearances. Moreover, Dhiraj et al [ 33 ] evaluated the Faster R-CNN [ 42 ], YOLOv2 [ 40 ] and Tiny YOLO [ 40 ] to detect contraband items such as shuriken , guns , knives , and razors from publicly available GRIMA X-ray Database (GDXray) [ 15 ].…”

Evaluation of a Dual Convolutional Neural Network Architecture for Object-wise Anomaly Detection in Cluttered X-ray Security Imagery

“…Prior work on appearance and semantic anomaly detection, has considered unique feature representation as a critical component for detection within cluttered X-ray imagery [26]. Early work on anomaly detection in X-ray security imagery [27], implements block-wise correlation analysis between two temporally aligned scanned X-ray images.…”

“…Early work on anomaly detection in X-ray security imagery [27], implements block-wise correlation analysis between two temporally aligned scanned X-ray images. More recently [28], anomalous X-ray items within freight containers have been detected using auto-encoder networks, and additionally via the use convolutional neural network (CNN) extracted features as a learned representation of normality across stream-ofcommerce parcel X-ray images [26]. Andrews et al [29] propose representational-learning for anomaly detection within cargo container imagery.…”

A New GAN-Based Anomaly Detection (GBAD) Approach for Multi-Threat Object Classification on Large-Scale X-Ray Security Images

“…While anomaly detection is commonly seen as a separate mode of operation in security inspection [13], in this paper we propose to use a novel approach to classify multiple threat objects in an x-ray image under different cases of class imbalance, in which we utilize a GAN-based anomaly detector coupled with a CNN and an SVM classifier. Inspired by the work in [11], we also adapt a Bi-GAN in our anomaly detector, but we extend it to work on higher dimensional data, i.e.…”