Researchers at the University of Granada in Spain, using machine learning techniques, have developed software that can detect a pistol in real time with high accuracy in video recordings or video feeds. The scientists' work can be found at arXiv.org, and a summary of it is given by Phys.org.
In connection with the growing global terrorist threat, work is being actively carried out today to create automated surveillance systems that could recognize people's faces, determine their emotional state, and also identify the objects they are carrying through video broadcasting in real time. It is believed that this will help prevent terrorist attacks.
Weapon detection systems existing today are metal detectors with adjustable sensitivity. These devices are considered effective, but require a person to pass through a special frame, which is sometimes difficult to organize. In addition, the detectors are not capable of detecting 3D-printed weapons.
The new software, created by researchers from the University of Granada, will allow the detection of small arms only by means of video surveillance. The weak side of the program, according to the developers, is only its inability to detect weapons hidden under clothing.
When creating the program, the scientists used a pre-trained neural network. Her object recognition training was based on ImageNet images, which includes about 1.3 million photographs of objects from about a thousand different classes. Accurate training in weapon recognition has already been carried out on 3,000 photographs of weapons prepared by researchers.
As a result, the researchers obtained a program that can identify weapons on video recordings with fairly high accuracy. The scientists checked the accuracy of the algorithm on recordings from YouTube, and they deliberately took recordings only in poor quality. The accuracy of weapon detection was 96.6 percent.
At the end of May this year, the South Wales police, thanks to a new automatic facial recognition system, detained a person suspected of a crime. Such a system was deployed in Cardiff in preparation for the UEFA Champions League final round, held on 3 June 2017 at Millennium Stadium.
Promotional video:
How exactly the facial recognition system used by the South Wales police is built has not been disclosed. It is only known that the equipment and software for it were supplied by the Japanese company NEC. It is only known for certain that the system operates on the basis of mobile minibuses with surveillance cameras. But both street cameras and police recorders can be connected to it.
Vasily Sychev