On the impact of using X-ray energy response imagery for object\n detection via Convolutional Neural Networks

2021/08/27 by Neelanjan Bhowmik, Bhowmik, Neelanjan, Yona Falinie A. Gaus +3 · 2 citations
Engineering · Computer Science · Dentistry · #Advanced X-ray and CT Imaging #Advanced Neural Network Applications #Dental Radiography and Imaging

paper · pdf · doi:10.48550/arxiv.2108.12505

Abstract

Automatic detection of prohibited items within complex and cluttered X-ray\nsecurity imagery is essential to maintaining transport security, where prior\nwork on automatic prohibited item detection focus primarily on pseudo-colour\n(rgb) X-ray imagery. In this work we study the impact of variant X-ray\nimagery, i.e., X-ray energy response (high, low) and effective-z compared to\nrgb, via the use of deep Convolutional Neural Networks (CNN) for the joint\nobject detection and segmentation task posed within X-ray baggage security\nscreening. We evaluate state-of-the-art CNN architectures (Mask R-CNN, YOLACT,\nCARAFE and Cascade Mask R-CNN) to explore the transferability of models trained\nwith such 'raw' variant imagery between the varying X-ray security scanners\nthat exhibits differing imaging geometries, image resolutions and material\ncolour profiles. Overall, we observe maximal detection performance using\nCARAFE, attributable to training using combination of rgb, high, low, and\neffective-z X-ray imagery, obtaining 0.7 mean Average Precision (mAP) for a six\nclass object detection problem. Our results also exhibit a remarkable degree of\ngeneralisation capability in terms of cross-scanner transferability (AP:\n0.835/0.611) for a one class object detection problem by combining rgb, high,\nlow, and effective-z imagery.\n

Cited by

• Can a Second-View Image Be a Language? Geometric and Semantic Cross-Modal Reasoning for X-ray Prohibited Item Detection
• Taming Generative Synthetic Data for X-ray Prohibited Item Detection

Related

• Operationalizing Convolutional Neural Network Architectures for Prohibited Object Detection in X-Ray Imagery
• Evaluating the Transferability and Adversarial Discrimination of Convolutional Neural Networks for Threat Object Detection and Classification within X-Ray Security Imagery
• Evaluation of a Dual Convolutional Neural Network Architecture for Object-wise Anomaly Detection in Cluttered X-ray Security Imagery
• Illicit item detection in X-ray images for security applications
• Illicit object detection in X-ray imaging using deep learning techniques: A comparative evaluation