Video Surveillance: Weapon detection

Weapon detection in video surveillance

Several recent statistics show that the rate of criminality caused by firearms is very concerning in many parts of the world, especially in countries where their possession is legally permitted. One way to reduce the violence generated by firearms is to detect their presence earlier so that the agents or guards can act. An innovative and effective solution in this context would be to provide surveillance cameras with intelligence.

The existent weapon detection systems are mainly metal detectors placed in the accesses in airports and at public events in closed places. Despite their robustness and necessity in certain places (airports, public buildings, …), these systems are very expensive, intrusive and not effective for the particularities of today’s world. There is currently a need for a system that can reinforce and improve existing systems and use them in a greater number of environments.

The final goal of this project is to develop a precise and robust intelligent system for the detection of weapons in videos especially suitable for the field of security.

The outstanding results are:

• A weapon detection model in videos based on Deep Learning (https://www.sciencedirect.com/science/article/pii/S0925231217308196).
• A cold-steel detection model in videos using preprocessing techniques and Deep Learning (https://www.sciencedirect.com/science/article/pii/S0925231218313365)
• An image fusion approach that minimizes the number of false positives in weapon detection in realistic video surveillance scenarios (https://www.sciencedirect.com/science/article/pii/S1566253518300393).

Period

January 2017- Present  

Researchers

Francisco Herrera, Siham Tabik, Roberto Olmos, Alberto Castillo, Francisco Pérez.

Awards

• Security Forum I+D+i award in 2017 to the innovation project UGR: “Real time video firearms detection system”(https://www.securityforum.es/en/awards/)
  
• Scientific recognition from MIT Technology Review to the arxiv version of the paper entitled “Automatic Handgun Detection Alarm in Videos Using Deep Learning” as one of the five most stimulating articles of the first week of March 2017 worldwide. (https://www.technologyreview.com/s/603786/the-best-of-the-physics-arxiv-week-ending-march-4-2017/)