0800-31-0700 for new subscribers
0800-31-0800 technical support

How do CCTV cameras recognise faces?

Home /

Blog

/

How do CCTV cameras recognise faces?

How do CCTV cameras recognise faces?

08.04.2024

Devices

846

In today's world, video surveillance systems are becoming increasingly popular. And one of their most common functions today is recognising human faces using CCTV cameras. Let's understand what principle they work on and where they can be useful.

Principle of Facial Recognition Technology

The use of FRT — a facial recognition technology based on the localisation of a certain person in an image — is becoming more and more widespread all over the world, in particular in China, the UK and Australia. This software solution identifies a person's face and subsequently identifies it almost accurately on the basis of the data available in the information database through the use of machine learning algorithms.

The algorithm of face recognition work can be based on such basic methods:

  1. Empirical approach. It was created in the era of the emergence of computer technology and consists of isolating an object from the general background and studying it in order to identify regular relationships.
  2. Method based on invariant data. It consists in identifying characteristic features of the shape of the face, the boundaries of its individual parts, establishing contrast and further combining all these parameters in order to verify the identity.
  3. Approach based on certain patterns embedded by developers in the neural network. In this case, each image segment is compared with existing standard templates, and this can be done from different angles in any scale dimension.

Face recognition algorithms, on the basis of which modern CCTV cameras work, may vary depending on the system. But in general, the process looks as follows:

Stage 1: Face Detection

Before a face can be recognised, it must first be detected. The camera captures a person's face regardless of whether he or she is alone or in a crowd. Modern technologies are able to recognise a person who is not even looking at the camera, but is within the permissible limits of head rotation.

Stage 2: Analysis of the received information

In order to carry out the analysis you will need photos of the face. Most systems do not work with 3D volumetric models, but use 2D format, as most databases have photos in 2D format. 3D volumetric images are rarely used, mainly for security reasons. The face of each person is analysed by at least 68 separate anthropometric points (in total their number can reach several thousand). There are also other landmarks (shape of cheekbones and chin, eye cut and distance between them, forehead height, etc.).


Step 3: Correcting distortions

To get the most accurate result possible, the recognised face should look full-face, i.e. directly into the camera. But this rarely happens, especially when it comes to recognising a person in a crowd. To improve the recognition quality, the system performs additional image transformation: it removes head rotation and tilt, as well as performs 3D reconstruction of the face from the 2D image.

Step 4: Digitalisation of the image

After the previous stage of analysis, the system uses a neural network to compile a specific digital facial code (feature vector or Face Print) that is unique to each person. In fact, it is a number that is a sum of facial characteristics: distances between reference points, texture of certain areas on the face, etc. The number of such characteristics can be many. There can be many such characteristics. The main rule is that they should describe the face regardless of extraneous factors: hairstyle, make-up, age changes.

Step 5: Image identification

The obtained digital code is compared with those contained in the database, where photos with digital identifiers are presented. The system searches for a match between the recorded photos and the samples available in the database. In this way, the possibility of obtaining additional information about a person increases


Applications of facial recognition systems

Facial recognition systems can solve a wide range of tasks in different fields:

Retail and entertainment venues

In shops, shopping centres, cameras with facial recognition are able to:

  • Determine the portrait of a typical customer and recognise his data.
  • Integrate with digital screens and show adverts to certain categories of visitors.
  • Determine the age of minors when buying a certain type of goods.
  • Identify thieves to prevent financial losses.
  • Identify consumer hotspots in-store.

Banking institutions

Here facial recognition cameras are installed for security control and identification of VIP clients.

Airports

Most of the world's airlines identify passengers after arrival and departure from the airport by scanning their faces for the purpose of further access to the aircraft. This procedure allows you not to carry a paper printed ticket, as information about the movement of each passenger who has passed the so-called "face control" is entered into an electronic database.

Stadiums

As we know, many countries around the world have problems with football fans. This is why some companies (e.g. Panasonic) embed special software into regular cameras that can identify blacklisted fans.

Educational institutions

To detect dangerous individuals, the FRT system is being tested by many American schools. It is also being used in Sweden to monitor students in class.

Pharmacies

The first facial recognition systems appeared in pharmacies in Shanghai in January 2020 to combat drug addicts and prevent pharmacists themselves from using drugs. And here China is ahead of the curve.

Exhibitions, installations and museums

Cameras (or exhibits) located in such places with facial recognition mechanics make it possible to determine the number of visitors at a particular exhibition stand, analyse their emotional reaction, etc.

Automotive industry

Japanese companies are introducing face recognition technologies into car production to detect signs of driver fatigue and prevent accidents.


Law enforcement and security agencies

For example, over 1,000 facial recognition cameras are already in operation across the UK to search for missing people, which carry out comparative analyses against a database of missing people. The facial recognition system in this country has been in operation for a long time and is now considered the most advanced in the world.

Mobile apps

Many app developers offer users the option to verify their selfies using facial recognition security when they log in to their apps.

The developed world has already made progress in the application of FRT technology in various industries. For example, China, which is one of the leaders in this industry, has already installed more than 170 million such cameras.

Among large global companies, facial recognition solution is also actively used by Amazon in its automated shops Amazon Go.

In Ukraine, facial recognition systems are also already being used both in public places and in the business sector. Cloud video surveillance from Maxnet, which is powered by software from Luxriot, also supports this technology.

Features of Luxriot video analytics system


Luxriot Vero Face Recognition is a biometric application designed for facial recognition.

Luxriot software is ideal for use at home or in a small office. The key features of this software are:

  • Works with thousands of IP and analog cameras.
  • Simultaneous multiple video streams processing.
  • Ultra-fast face detection and matching against the database using DNN.
  • Flexible rule configuration for triggered faces.


To order a cloud video surveillance service for home or office based on Luxriot, please contact the manager of the company Maxnet in a convenient way for you.

Comments

0

Еще комментарии