An Automatic Hybrid Approach to Detect Concealed Weapons Using Deep Learning

Detecting concealed weapon underneath a person's clothing is important for public safety in places such as airports. The success rate of our proposed approach in detecting concealed weapons based on fused images is higher than that of the bag-of-features approach. Our approach uses a traditional discrete wavelet transform with hybrid bag-of-words to obtain fused imagery. We then utilized convolutional neural network (CNN) with a pre-trained CNN model using the CNN features of the fused image to train a multiclass SVM classifier. Our approach works well with X-ray images. The experimental results indicate the efficiency of the proposed hybrid approach.