ANALYSIS OF METHODS FOR DETECTING FACES IN AN IMAGE

Abstract

In this article, computer vision is considered as modern technology of automatic processing of graphic images, and the relationship between the terms “computer vision” and “machine vision” is investigated. A diagram of a typical computer vision system is given and the possibility of using a system based on an artificial neural network for image analysis is considered. The article analyses the current situation with the use of computer vision systems and the possibility of its application. This article presents face recognition algorithms for existing categories, including: empirical method; feature method – invariant feature; use the template specified by the developer for identification; study the method of detecting the system by external signs. The empirical method of “top-down knowledge-based methods” involves creating an algorithm that implements a set of rules that image segments must satisfy in order to be recognized as faces. Feature-invariant approaches (Feature-invariant approaches) based on bottom-up knowledge constitute the second group of face detection methods. The methods of this group have the ability to recognize faces in different places as an advantage. Use the template set by the developer for identification (template matching method). Templates define specific standard images of face images, for example, describing the attributes of different areas of the face and their possible mutual positions. A method for detecting faces by external signs (a method for performing the training stage of the system by processing test images). The image (or its fragments) is somehow assigned a calculated feature vector, which is used to classify the image into two categories – human face/non-human face.