Artificial intelligence beyond high-end video surveillance applications

Uri Guterman, director of product and marketing of Hanwha Techwin Europe, explica por qué la inteligencia artificial con aprendizaje profundo ya no es una tecnología emergente que solo se usa en videovigilancia.

Unless you have recently participated in Video surveillance involving artificial intelligence (The) With deep learning (deep learning), you may have the impression that the technology is too expensive to deploy in other non-high-end applications.

With the introduction of a new generation of affordable cameras that already integrate AI with deep learning things have changed a lot. However, the jargon associated with technology, such as artificial neural networks and machine learning, could give the impression that the possibilities of technology go far beyond what most users s need to achieve the maximum benefit of their video surveillance solutions.

Removal of false alarms

This is also far from the case, since most of the situations in which AI cameras could be installed will be those in which there is a basic need to solve the old problem of false alarms, very common in the electronic security industry for decades.

Simply put, AI-based video analytics with deep learning ignores video noise, the leaves of trees moving, clouds moving and animals.

In short, anything that could normally be the cause of false alarms when using sensors or standard motion detection technology to detect activity, as they have not been designed for that purpose.

This higher level of performance of AI-based video analytics with deep learning means control room operators and security personnel can focus on responding to real incidents and emergencies, and not waste time and effort on false alarms.

In addition to extreme precision, deep learning it also allows operators to search for specific features and attributes, including a person's age group and sex, if you are wearing glasses, hat or a purse.

Easy to use and configure

With all the intelligent elements already integrated in the cameras with AI there is nothing complicated in the installation, configuration and use of artificial intelligence with deep learning. Like this, system integrators will be able to apply this technology in virtually any video surveillance project.

The AI is ready to work as soon as the camera is installed, although it offers the opportunity to be customized to meet the operational requirements of the end user and it is not necessary for the end user to have a deep technical level. However, we will give a technological perspective.

Let's start with deep learning, which is part of machine learning and is a way to achieve AI by training a machine to perform tasks based on a lot of examples.

To achieve this, deep learning uses deep artificial neural networks, or multilayered, which are essentially mathematical models inspired by the human brain.

The fact that they are deep makes them very suitable for solving complex problems, such as identifying and recognizing objects and events in raw video streams, with very high accuracy.

As an example, to correctly establish a person's gender requires expert R&D engineers to design, teach and validate a deep learning network that, during the training stage, uses a database of millions of properly selected faces, each of which is labeled with its true known genus.

After several days of training by our engineers, the neural network is ready to go live and is likely to have an approximate accuracy of 98%, which is more or less the same as the ability of humans to do the same task.

AI techniques with deep learning offer far superior performance compared to more traditional video analytics. the latter generally use motion detection and are not advanced enough to detect static objects (as parked vehicles) or troubleshoot video noise, light pollution from headlights or moving shadows, which are the cause of false alarms.

Analytics performance is equally impressive in environments where there are fast or very busy movements, improving the search for expert evidence and speeding up investigations.

For these and other reasons, perhaps it is inevitable that AI with deep learning gradually replace traditional video analysis in most applications, and particularly those that suffer more false detections.

AI with deep learning is particularly suitable for applications that require a greater degree of sophistication than traditional video analytics.

For example, enables businesses to capture and apply business intelligence for people's age and gender, so that it is possible to analyze the demographics of customers on an individual basis and, in doing so, gain a greater understanding of customer behavior and buying patterns.

It should be noted that artificial intelligence with deep learning has made a valuable contribution over the past year, as it has been at the heart of mask detection apps, distance measurement and occupancy monitoring.

In addition to helping to combat criminal activity, there are countless ways in which AI video analysis with deep learning integrated into cameras can help businesses improve productivity and operate safely in a pandemic-affected world.

With the recent availability of inexpensive cameras equipped with this technology, users can now expect a high return on investment, regardless of how they leverage technology.

Uri Guterman

Product and Marketing Director of Hanwha Techwin Europe

 

 

 

Other articles on Hanwha Techwin