TUSUR: The system of face recognition based on the surveillance camera

For conversion of the standard IP camera to an intelligent system of video surveillance developers of the Tomsk state university of management systems and radio electronics (TSUCSR) create the software package consisting of several modules. It it was announced in June, 2015. He will allow a system to define and distinguish in real time a human face in the movement, from different foreshortenings.

Authors of work - the Kamensk Andrey (the 5th rate of RTF, specification department, specialty the audiovisual equipment), Malanin Maxim (the 5th rate of RTF, specification department, specialty the Audiovisual equipment), Kapustin Vyacheslav (the graduate student, specification department). The research supervisor – Kuryachy Mikhail Ivanovich, PhD in Technological Sciences, the associate professor of the specification.

Development, according to her creators, will be useful to automatic fixing of visit of occupations by students, school students and also to simplification of a gating system at the enterprises.

Important advantage of a "smart" system which is created in TUSUR, - that it can work with any IP cameras which are widely presented at the market, unlike analogs which require installation of special expensive cameras. In addition, for successful recognition it will not be required to set special barriers which would direct people directly to the camera, it will not be necessary to stop specially in front of the camera.

Developers – group of students and graduate students of department of television and management of TSUCSR – explain that the IP camera will work in the normal mode, i.e. to take a video flow and to transfer him to the server where the special software module will select a human face in a video flow, another – will process and distinguishes it. For more exact recognition a system is created with self-training function. It gains a large amount of signs of each person, fixes changes, independently classifies information relating to the specific person. And on the created images will be able to distinguish persons from any foreshortening.

Authors of development are going to create the general system, a packet of modules which can be applied to any IP cameras. Usually, in addition to purchase of special cameras, for work of the similar systems modules of identification of the person in a video flow and recognitions of the person in the base having restrictions for the number of persons which it can save are separately purchased (from 1 to 5 thousand). Tusurovsky development of such restrictions does not assume, a system will be able to keep enough images of each person for better recognition.

For setup and debugging of a system in TUSUR the educational laboratory in which lecture and practical classes are is equipped. As developers note, a system at first will be tested on small audience for 25 seats, and after it it will be possible to set on large, stream audience. In addition to recognition, a system will fix also automatically students in the electronic journal.