Cascaded structure tensor for robust baggage threat detection

In the last two decades, baggage scanning has become one of the prime aviation security concerns worldwide. Manual screening of the baggage items is tedious and an error-prone process that also compromises privacy. Hence, many researchers have developed X-ray imagery-based autonomous systems to address these shortcomings. This paper presents a cascaded structure tensor framework that can automatically detect suspicious objects from the baggage X-ray scans under extreme class imbalance and irrespective of the baggage clutter. The proposed framework is unique as it intelligently extracts each object by iteratively picking its contour-based transitional information from different orientations and uses only a single feed-forward convolutional neural network for the recognition. The proposed framework has been rigorously evaluated on publicly available GDXray and SIXray datasets for detecting the highly cluttered and overlapping suspicious items, where it achieved the mean average precision score of 0.9343 and 0.9595, respectively, across both datasets, outperforming state-of-the-art works by 1.94% on the GDXray, and 8.21% on the SIXray. Furthermore, the proposed framework gives the best trade-off between detection performance and efficiency.