An international project develops an intelligent video surveillance system for large cities

The main objective of the HuSIMS project, in which the University of Valladolid participates, is to overcome the existing limitations in current video surveillance solutions and generate a platform that allows increasing security in public places in large cities..

Deploying a video surveillance system in cities with millions of people is a complicated challenge. Added to the costs of cameras and infrastructure are those of human operators, since a large number of personnel are needed to monitor what happens in each camera. In order to take advantage of the possibilities offered by ICT in this sense, ten entities of Israel, Türkiye, Korea and Spain, among them the University of Valladolid as the only participating university, have started the HuSIMS project. The work is part of the European Commission's Eureka-Celtic initiative, whose objective is to promote collaborative R&D to increase competitiveness in telecommunications.

Javier Aguiar, CIT Group coordinator (Comunication and Information Technologies) from the University of Valladolid, has specified that the HuSIMS project aims to design a network of video surveillance cameras capable of intelligently identifying when an emergency situation is occurring in the image., from a traffic accident to a fire. “These are monitoring and video surveillance systems with massive sensor networks to control large metropolitan areas intelligently., that is to say, without the need for a human operator to constantly monitor what is happening in the cameras”, has pointed out.

This way, Thousands of cameras could be controlled with few operators, since they would only receive alerts in anomalous situations. “The cameras don't even take the images, but rather they model them through mathematical models, and those parameters that are capturing (the position, size or speed of objects), being data, they take up much less bandwidth, which also reduces the transmission cost, because they are wireless cameras that send all the information to a central unit”, details. So, duck, Camera costs are also reduced. Two leading Israeli companies in this type of devices participate in the project., “cameras that, being wireless, avoid infrastructure costs and, by not directly processing the information, but they send it to a centralized unit, allow low cost.

Training

As explained by Javier Aguiar, These video surveillance systems based on artificial intelligence require “training” time to distinguish anomalous behavior from those that are not.. “Once you prepare the systems, the operator no longer has to be aware of the cameras, only from the alarms they generate”, remember, very useful information for emergency teams.

Researchers are currently conducting tests around different anomalous situations., like traffic accidents. “Once we train the cameras so they know what they are monitoring, in this case road safety, Then it is relatively simple, through mathematical models, to know if what happens is normal or if the parameters taken are outside of what was expected., such as a car driving in the opposite direction.”, Aguiar has stressed, who pointed out that it is about using artificial intelligence “in a field in which it had not yet been applied”.

The fact of “adapting” video surveillance to different situations means, in the opinion of the expert, an important added value. “One of the most powerful things about these new technologies is that the core is common and can be applied to different uses.”, so we also work on vandalism or environmental issues”, highlights. In this last field, satellite cameras are being used to control forest fires. “The only thing you have to do is change the rules of artificial intelligence so that it detects some anomalous cases or others”.

The concern on the part of the European Commission in this line of investigation has increased in recent years to avoid cases of terrorism.. In this sense, Aguiar advances that there are already security companies interested in the results of this project, which will conclude in 2013.

Detection of anomalous patterns

The work carried out by the University of Valladolid within the framework of the project focuses, according to the coordinator of the CIT group, in the artificial intelligence part. “Once the cameras capture and model the information, They pass us the parameters of the objects. “We work on the artificial intelligence part to detect those anomalous situations or patterns within which the cameras are passing us.”, has pointed out.

To implement this intelligent behavior, one of the strategies used is semantic analysis. A first analysis of moving objects and their trajectories in the video signal allows identifying entities with meaning such as roads, sidewalks, doors, pedestrians or vehicles. In a second stage, these objects are incorporated into a semantic knowledge model and the normal and anomalous behavior of these actors is characterized.. This allows the system to operate at a level of abstraction similar to that of human knowledge., as the researchers explain, who hope to have a first demonstrator of the system next year.