Face recognition system with the help of video surveillance systems. Face Search Algorithm

Software and hardware of modern integrated security systems are able to solve problems of any complexity on all kinds of industrial, social and domestic objects. Very important tools of security systems are video surveillance systems, and the requirements for the functionality of the segment are steadily growing.

Integrated security systems

A single platform of integrated security systems includes modules of security and fire equipment, control and access control, video surveillance or security television (COT). The functions of the latter until recently were limited to video monitoring and recording of the situation at the site and adjacent territory, archiving and storage of data. Classical video systems have a number of significant drawbacks:

• Human factor. Inefficient operation of the operator when broadcasting a large amount of information.
• Impossibility of surgical intervention, untimely analysis.
• Significant time costs for finding and identifying the event.

The development of digital technologies led to the creation of "smart" automated systems.

Strength in Intelligence

The basic principle of an intelligent video surveillance system is video analytics - a technology based on methods and algorithms for image recognition and image processing, automated data collection as a result of video stream analysis. Such equipment without human intervention is able to detect and track in real time the set goals (car, group of people), potentially dangerous situations (smoke, fire, unauthorized interference with the operation of video cameras), programmed events and timely issue an alarm. Due to the filtering of video data of no interest, the load on communication channels and the archive database is significantly reduced.

The most popular means of video analytics is the system of face recognition. Depending on the functions performed and the tasks assigned, certain requirements are imposed on the equipment.

Firmware

For effective operation of the system, several types of IP cameras with different operational characteristics are used. Detection of the object in the controlled territory is fixed by panoramic view cameras with a resolution of 1 Mp and a focal length of 1 mm and are guided by scanning devices. These are more advanced cameras (from 2Mp, from 2mm), which produce recognition by simple techniques (3-4 parameters). To identify the object use cameras with good image quality, sufficient for the application of complex algorithms (from 5 Mp, 8-12 mm). The most popular software for face recognition "Face Intellect" (developer - the company House Control), Face director (the company "Synesis" and VOCORD FaceControl (VOCORD) demonstrate:

• High probability of identifying an object (up to 99%).
• Support for a wide range of camera rotation angles.
• The possibility of separating people even in a dense pedestrian mass.
• The variability of the preparation of analytical reports.

Basics of pattern recognition

Any biometric recognition systems are based on the identification of the correspondence of the physiological characteristics of the person being read to a specific predefined pattern. Scanning takes place in real time. The IP camera broadcasts the video stream to the terminal, and the face recognition system determines the correspondence of the image to the photos stored in the database. There are two main methods. The first is based on static principles: the results of processing biometric parameters create an electronic sample in the form of a unique number corresponding to a particular person. The second method models the "human" approach and is characterized by self-learning and robustness. Identification of a person according to the video image takes into account age-related changes and other factors (presence of a headdress, a beard or mustache, glasses). This technology allows you to work even with old photos and, if necessary, with X-rays.

Face Search Algorithm

The most common method of detecting faces is using Haar cascades (sets of masks). The mask is a rectangular window with a different combination of white and black segments.

The mechanism of the program is as follows: the video frame is covered with a set of masks, and according to the results of the convolution (counting the pixels caught in the white and black sectors) a difference is calculated, compared with a certain threshold value.

To improve the work of the classifier, positive (frames, where people's faces are present) and negative (without those) training samples are created. In the first case, the result of convolution is above the threshold value, in the second case it is lower. The detector of persons with an allowable error determines the sum of the convolutions of all cascades and, if the threshold is exceeded, signals the presence of persons in the frame.

Recognition technologies

After detection and localization, the brightness and geometric alignment of the image occurs in the preliminary stage. Further actions - calculation of features and identification - can be carried out by various methods.

When scanning a full face in a room with excellent lighting, good results are demonstrated by algorithms that work with 2D images. Analyzing unique points and distances between them, the face recognition system determines the fact of identification by the coefficients of the difference between the "live" snapshot and the registered template.

Three-dimensional technologies are resistant to changes in the light flux, the permissible deviation from the front angle is up to 45 degrees. Here, not only points and lines are subjected to analysis, but also the properties of surfaces (curvature, profile), the metric of distances between them. For the operation of such algorithms, the maximum quality of video recording with a frequency of up to 200 frames per second is required. The basis of the system consists of stereo cameras with a matrix of 5 megapixels, high optical resolution and minimized synchronization error. In addition, they are connected by a special timing cable to transmit the sync pulses.

The state of the modern systems market

The first systems of biometric control, because of their high cost, were developed only for state military facilities and only in the mid-1990s they became available to commercial organizations. The rapid development of technology and microprocessor technology has allowed to increase the accuracy of systems and expand the scope of their application. In the market of our country leading positions belong to American and West European security system manufacturers. The leader of sales is the equipment of corporations ZN Vision Technologies and Visionics. The most promising among Russian developers are research and products of Vocord, NTechLab, Soling, VizhnLabs and MDG groups, which, among other things, are also engaged in adapting foreign complexes to Russian conditions.

Computer Face Control

The most extensive field of application of contactless identification is the fight against terrorism and crime. The image of the criminal's face is stored in the database. In places of mass congestion of people (airports, railway stations, shopping mall, sports facilities), a real-time stream of people is being filmed for the identification of wanted persons.

The next sphere is access control systems: the sample of the photo image on the electronic pass is compared with the model obtained as a result of data processing from video cameras. The procedure takes place instantly, without requiring any additional actions (as opposed to scanning the retina of the eye or dactyloscopy).

Another rapidly developing industry is marketing. An interactive billboard, having scanned the person's face, determines his gender and age, visualizes only those advertisements that will potentially be of interest to the client.

Trends and development prospects

The systems of recognition of persons in the banking sector are very in demand. As a result of last year, after the installation of 50,000 intelligent video cameras in their offices, the Post Bank Bank managed to save millions of rubles by preventing fraud in the credit and payment segments. Experts assert that by 2021 the necessary infrastructure network will be created and any operations in ATMs will become possible only after the biometric identification of the client's face.

In the next decade high technologies will allow to open a network of shops full of self-service: the buyer passes in front of the storefronts, chooses the goods they like and leaves. The system of recognition of faces and images will determine the identity of the buyer, purchase and will debit his account the necessary amount.

Work is underway to create systems for recognizing the psychoemotional state. The analysis of human emotions will be in demand in multimedia spheres: animation, cinematography, the industry of creation of computer games.