Concealed Object Detection and Localization in Millimetre Wave Passengers’ Scans

The exponential growth in air travel has heightened airport security risks, making traditional manual screening methods increasingly inefficient. To address this challenge, we propose an automated system for detecting and localizing concealed prohibited objects in millimeter-wave images of passengers. Our approach leverages image processing and data mining techniques to enhance detection accuracy and efficiency. In our method, each millimeter-wave image undergoes preprocessing to filter out noise and artifacts. The images are then segmented into zones, each treated as an individual image. Zero-centering and normalization are applied to these zones to optimize the performance of the neural network. The dataset is divided into training and testing sets, with the training set shuffled to improve learning. The neural network is trained on this data to predict the presence of potential threats. This paper presents the methodology of our solution and discusses how it addresses the challenges of passenger scanning and false alarms. Preliminary results indicate that our system has the potential to significantly improve detection rates while reducing unnecessary alerts, thereby enhancing overall airport security efficiency.