Forbidden objects, X-ray scan images, Preprocessing, Segmentation, Feature extraction, Classification, Handguns, Customs border security, Airport security

Detection of forbidden objects in x-ray scan images has become an important issue for customs border and airport security. Most border screening depends on the manual detection of possible forbidden objects by human experts.In this thesis, we present a system for detection of possible forbidden objects in x-ray scan images with minimal amounts of missing object (false negatives) and false alarms (false positives). Firstly, pre-processing steps are applied to obtain a clearer image. Then, segmentation is used to locate the potential objects in images. Feature extraction with two algorithms, local binary pattern and histogram oriented gradients, and classification with support vector machine are next steps in the system.The system is tested using handguns as the forbidden objects in question. The experimental results show that the system can effectively detect the handguns in x-ray scan images with minimal amounts of missing objects automatically.