Explosive Detection from X-ray Images

 

An automated detection system is being developed in our lab to detect explosives and other contraband, such as guns and knives, by imaging the complete contents of the parcel in three dimensions. Our system uses technology derived from elastic X-ray scatter approach and computer vision methodology to locate and characterized threatening object according to their unique diffraction profiles. After data acquisition from the X-ray system, the detected signal will be evaluated by the computer. At first, the data is enhanced in a form that it is suitable for computer analysis by utilizing adaptive nonlinear noise removal filter. Data segmentation is performed to isolate explosive candidate regions for further analysis using fuzzy cluster analysis. Features of the suspect regions are used to categorize as explosive or non-explosive material using a novel neural network based classification algorithm. The preliminary analysis of our system shows that we can detect plastic explosive, industrial explosive and even home-made explosives.