Improving Threat Object Recognition for X-Ray Baggage Screening Using Distraction Removal Techniques

X-ray security screening plays an important role in maintaining safety and security at railway stations, airports, and in the transportation system. This process involves the inspection and examination of luggage and personal belongings to detect threat items and explosive items. The complexity of luggage and baggage scanning can indeed make threat detection challenging. This paper proposes a method to improve visual search tasks in baggage security screening by utilizing automatic instance segmentation act as a diagnostic aid, drawing the screener's attention towards the threat items. This involves an inpainting method using regionfill to automatically detect and remove benign objects from the image. The average confidence score of the detected item after the regionfill approach is increased from 98% to 99.5% for threat items (gun, knife, shuriken, blade). By integrating an automatic segmentation model and inpainting methods, this work offers a promising avenue for improving the effectiveness of X-ray screening procedures in identifying potential threats.