Automatic threat object classification based on extracted features from electromagnetic imaging system

The detection of concealed weapons is one of the biggest challenges facing homeland security. It has been shown that each weapon can have a unique fingerprint, which is an electromagnetic signal determined by its size, shape, and physical composition. Extracting the signature of each weapon is one of the major tasks of any detection system. In this paper, feature extraction of a new metal detector signal is conducted using a Wavelet and Fourier Transform. These features are used to classify two different groups of threat objects. Artificial Neural Network (ANN) and Support Vector Machines (SVM) classification techniques are used to classify the metal objects towards automatic threat object detection and classification. Promising classification accuracy rates are obtained from using individual and combined features.