Formal Grammar Theory in Recognition Methods of Unknown Objects
- PDF / 543,728 Bytes
- 11 Pages / 612 x 792 pts (letter) Page_size
- 96 Downloads / 212 Views
RMATION ANALYSIS
Formal Grammar Theory in Recognition Methods of Unknown Objects N. I. Sidnyaeva, *, Yu. I. Butenkoa, **, and E. E. Bolotovaa, *** a
Bauman Moscow State Technical University (National Research University), Moscow, 105005 Russia *e-mail: [email protected] **e-mail: [email protected] ***e-mail: [email protected] Received May 24, 2020
Abstract—Questions on the formation of contextual grammars that describe both the structural information of an image and the interaction of images in a complex scenario have been considered. The use of a multilevel grammar is proposed, including the task of parsing a sequence of images, as well as the task of parsing objects for various purposes, when the nature of the source is not clear. It is shown that the formation of a grammar that describes both the structural information of an image and the interaction of images is associated with the need to develop an algorithm for recovering the grammar from a given set of dynamic images that represent a training sample. Some basic provisions inherent in structural methods for describing and recognizing a scene are presented. Keywords: recognition, grammar, syntactic method, evaluation, solution, image, learning DOI: 10.3103/S000510552004007X
INTRODUCTION Computer vision is one of the most popular areas in the field of intelligent systems, including complex recognition algorithms and complex mechanisms of deep machine learning. Computer vision is based on the reproduction of the mechanisms of the complex system of human vision, which allows one to obtain meaningful information from images, videos, and other visual aids, as well as taking the necessary measures or making recommendations based on the received information. Until recently, computer vision worked only within limited capabilities, but due to advances in artificial intelligence and innovations in deep learning, this area has been actively developed in recent years and has largely surpassed humans in performing some tasks related to the detection and recognition of objects [1–3]. One of the driving factors in the development of computer vision is the amount of data that is used to train and improve such systems [4–7]. The problems that arose in solving the problems of image recognition led to the development of various algorithms, one of which is the structural one, which is called the linguistic or syntactic method [4, 5]. Its peculiarity lies in the fact that the a priori descriptions of classes are structural descriptions, that is, formal constructions in which the principle of taking the hierarchy of the structure of an object into account and taking the relationships that exist between individual elements of this hierarchy within the same levels and
between them into account is consistently carried out. Structured recognition is used for a wide variety of tasks: character recognition, speech and fingerprint recognition, scene analysis, and chemical and biological structures [6]. In many cases, a posteriori information about the recognized objects is
Data Loading...
