Accelerated pistols recognition by using a GPU device

In most big cities, firearm assault is a common crime. Some state of the art research aim to recognize firearms once they were fired. However, to prevent this type of criminal behavior it is necessary to detect firearms in real time, before they are fired, and maintaining at minimum false alarms. In this paper, we propose a method to detect hand guns by using its shape and real dimensions. The proposed method is based on statistics invariant moments, image processing algorithms and a neural network classifier. To reduce processing time a GPU processor is utilized. Computer simulations show promising results.