Firearm Detection Through Component Decomposition: practical application of AI powered surveillance system
Abstract of the Thesis:
Firearms involving violence, terrorist attacks, and school shootings usually happen in public places with large crowds to cause the most damage possible and get the most attention. Surveillance cameras are a widespread and powerful tool, but their effect in preventing crime is far from perfect due to either environmental conditions or human factors. A surveillance system that can detect, analyze, and react to a criminal activity involving firearms may help save lives and reduce crime. This research focuses on detecting weapons within images or video frames with the help of semantic decomposition; a proposed method that divides the problem of detecting and locating a weapon into a set of smaller problems concerned with the individual component parts of a weapon. The developed system was tested and evaluated. The results showed that the achieved performance is on par with traditional machine learning solutions, such as object classification with a convolutional neural net, but also possesses additional qualities such as modular design, flexibility, and low complexity. With a classification accuracy over 90%, the system was able to successfully detect the entire firearm as well as unusual combinations of different parts of the weapon.