Information Control Systems and Technologies
The object of computerization is a video surveillance system. Video surveillance is successfully used in large areas crowds of people – in airports, various supermarkets, Parking lots. In this case, the systems allow to produce round-the-clock monitoring of the situation and instantly inform about unforeseen situations. Thanks to this the use of video surveillance systems significantly reduces the possibility of dangerous to human life situations'.
In connection with the above, it can be concluded that this is one of the most effective technical means of ensuring security, allowing you to quickly record the fact of the Commission of a wrongful act.
At the moment, there are many different methods and algorithms that allow you to process the image and get information about the persons present on it. To select the optimal solution in the conditions of the problem to be solved facial recognition.
Facial recognition is a task that a person performs effortlessly several times a day. This is one of key biometric technologies. Face recognition has a number of advantages over other biometric technology: it is natural, accessible and easy to use. Based on this, the issue of improvement and modernization of this type of recognition systems is relevant.
The aim of this work is to increase the level of security through the integration of personal identification and through portrait expertise in video surveillance systems.
For this purpose it is necessary : to investigate the subject area, to analyze the existing methods of solving such problems. problems, highlight their advantages and disadvantages, choose the most promising methods of solving this problem, analyze the results of their application and choose the best one.
Formed a number of tasks to achieve this goal:
The task of identification of the person is rather difficult, especially for tasks without restrictions, when the point review, lighting, facial expression, fence, accessories, etc. can change significantly. Face like a three-dimensional object must be identified based on its two-dimensional image. The image of the face may change when you change the lighting, poses, expressions, etc. a Typical face recognition system consists of four stages:
The results of the face recognition problem depend significantly on the extracted features and classification methods. A table view of the task.
| Task | Source. data | Result |
|---|---|---|
| Detection | Image | Making a decision on the presence( and possibly the number) of persons in the image, determining their position |
| Recognition | Fragment with one face | Candidates from the existing database, the key features of which are close to the data of the current fragment. |
When building an automatic face detection system, it is necessary to take into account the following features that complicate task:
Analysis of facial recognition methods.
Almost all methods are based on the use of local or global facial features.
local features, the algorithm allocates individual parts of the face (such as eyes, nose, mouth, etc.) and already
on them allocates
or recognizes face. When using global features, the algorithm operates on the whole face.
Methods based on local features, one way or another, localize characteristic areas of individuals and already on their basis perform further processing. The disadvantages of this group are as follows:
Popular representatives of these methods are the method of comparison of elastic graphs and neural networks.
Methods based on global features include linear discriminant analysis, principal component analysis, independent factor analysis. The basis of all methods based on the analysis of global features is, to build a certain partition of multidimensional space, dividing the areas belonging to different people. It this will determine which of the N image classes the new image belongs to, i.e. which of the N people depicted on it. These include:
| Method | Recognition accuracy | Effect of facial expressions on recognition accuracy | Recognition time | Computational complexity |
|---|---|---|---|---|
| Flexible comparison on graphs | ~90% | low | ++ | ++ |
| Principal component | ~90% | high | + | + |
| Neural network | >90% | low | + | + |
| Viola-Jones | >90% | low | + | + |
taking into account the above, it seems that the creation of hybrid methods that use advantages and leveling disadvantages of the above various special approaches. One of options such a hybrid is the use of the viola-Jones method to recognize present faces in a frame, and the given method refers to a class of methods based on the simulation of images of persons. Using this algorithm, the subsystem will get parts of the image that contains only a set of entities then the task is executed recognition, for which it is advisable to use the principal component method.
Based on the information obtained, a solution is proposed that combines the use of the viola-Jones method for face detection in the frame, because this method has a high recognition rate and a low probability of false operations. The method of principal components is proposed to be used as a recognition method.
This solution is based on the fact that for the majority of modern systems of automatic face recognition the main the task is to compare a given image with a set of images of faces from the database. Characteristics automatic face recognition systems in this case are evaluated by determining the probability of failure in recognition and mistaken recognition. And the proposed option solves these basic problems, as it allows store information in a convenient format (feature set), and has the ability to quickly and accurately determine the presence of persons in the frame by using a cascade of features.
The direction of further research is the problem of choosing the most suitable frame from the video stream for its subsequent processing.
At the time of writing this essay master's work is not yet completed. Estimated completion date: may 2019 the Full text of the work, as well as materials on the topic can be obtained from the author or his supervisor after indicated date.