Проектирование систем безопасности с идентификацией пользователя по лицу Реферат

Human face is one of the most informative means of communication in our societal life. Unlike, face recognition by humans to comprehend their peers possess a natural phenomenon, but recognizing facial geometry through machine is still a challenging problem. Face recognition is the task of recognizing an individual using

digital facial image.

The purpose of the work is to present an overview and a general classification of face recognition methods along with their pros and cons. 

FR is biometric application in which an individual face is identified in a digital image by analyzing and comparing patterns. 

Among many biometric traits, face is considered as one of the major biometric. Although it is quite feasible to access face images, one needs to address many challenging problems such as face occlusion, blurry image, poor illumination, both

non-linear and linear transformations of face images, ageing etc. To efficiently handle these challenges, researchers have come up with robust algorithms to extract facial

feature as well as several mathematical models for preprocessing such features for better classification.

Diabetic macular edema (DME) is another major reason for low vision and blindness mainly in diabetic population. Visual loss from DME is five times more than that from proliferative diabetic retinopathy (PDR). Identification of macular area of human beings plays an important role in the detection and diagnosis of many eye

diseases for ophthalmologists. Macular changes can happen during the progress of diabetes mainly associated with non-proliferative diabetic retinopathy (NPDR) or proliferative diabetic retinopathy (PDR).

The Face Recognition is growing as a major research area because of the broad choice of applications in the fields of commercial and law enforcement. So it leads and encourages the researchers for continuous research in this area.

Face Recognition has grown to be major study areas of computer visualization and applied pattern recognition. Principle of Face Recognition is lying on input facial images, collected in electromagnetic range to their identification and verification, but beyond that, it has gain the mature level with some limitations. 

A number of methods have been developed by different researchers to minimize the difficulties on the imagery, collected in Visible Spectrum. The problem of illumination is minimized by using the two techniques called as Image Processing Filters and Statistical Facial Models. 

Due to the support to the invariance in the changes of poses, the infrared spectrum has achieved a great consideration in the research domain. In Face Recognition, the images captured by cameras in infrared spectrum have more characteristics than Visible Spectrum. The infrared images can be captured in different type of lighting environment; even in the complete dark night, which show the strength

of infrared images to represents the facial variances for Face Recognition. Another characteristic of infrared images is the effect of absorption and scattering on the energy, less as compared to the Visible Spectrum.

Researchers divided IRS into four sub-bands namely NIR, SWIR,

MWIR and LWIR. These sub-bands contain the different frequencies that value due to the solar spectrum and divided by absorption lines of the atmospheric gasses.

IR based FR techniques are developed due to the limitations in the visible VS and leads to a new research in the IR field for which several complex IR specific models are being proposed in which the domain specific properties of

the data are much explored. The IR images have a great significance in the field of FR to extract important biometric features, facial expression, pose changes etc., which demonstrate a high level of distinctiveness and affect the FR system.

Automatic facial recognition can be seen as a pattern recognition problem, which is very challenging due to its nonlinearity. It can be viewed as a template matching problem, where matching has to be performed in a high dimensional space. A face recognition system (FRS) is typically designed to measure the similarity between probe and gallery facial images.

Face recognition systems are also looking at ways to apply the latest advances in FRT to uncontrolled environments, where state-of-the-art face detectors can achieve about 50-70% detection rate, with about 0.5-3% of the detected faces being false positives. The matching process in a FRS generates the similarity score from some probe image for each image in the gallery and taking the decision on the closest match in contrary to a human that examines the top-most matches.

A number of face recognition methods have been developed during the past decades. We can classify these methods in two groups that include appearance-based methods and model based methods. Former methods use holistic texture features that are applied to either whole-face or specific regions in a face image whereas latter

methods employ shape and texture of the face.

Face recognition problem can be considered as a space-searching problem in combination with machine learning problem. The most popular methods of appearance-based face recognition are: Principal component analysis (PCA), Independent component analysis (ICA) and Linear discriminant analysis (LDA). The PCA finds a set of the most representative projection vectors such that the projected samples retain most information about original samples. The ICA captures both second and higher order statistics and projects the input data onto the

basis vectors that are statistically independent as possible. The LDA uses the class information and finds a set of vectors that maximize the between-class scatter while minimizing the within-class scatter.

Cognitive Load Theory (CLT) uses interactions between information structures and knowledge of human cognition to determine instructional design. The theory’s initial development in the early 1980s provided instruction that differed from the prevailing orthodoxies of the time. An emphasis on instruction designed to reduce unnecessary or extraneous cognitive load resulted in, for example, a recommendation to provide learners with many worked examples rather than problems to solve. By considering relations between working memory and long-term memory, it was possible to structure both worked examples and related instruction to further cognitive load. A strict insistence that all recommendations be directly tested for effectiveness using controlled experimental designs also provided a point of departure from the prevailing orthodoxies, which tended to recommend instruction based on process models alone without testing comparative instructional effectiveness. 

According to CLT, there are three types of load:

1) Intrinsic load depends on the level of complexity of the topic;

2) Extraneous load depends on any activity that is not related to authoring the content;

3) Germane load is related to how effectively the content can be organized.

ADDIE is an instructional design model that helps the designers complete the task effectively in a systematic manner. ADDIE model comprises of five core phases such as: Analyses, Design, Development, Integration and Evaluation. The system adheres to this instructional design model.