ISSN 2079-3537      

 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                                                                                             

Scientific Visualization, 2018, volume 10, number 5, pages 57 - 85, DOI: 10.26583/sv.10.5.05

The image analysis of the geometric bodies with supplemental interactive block for training new knowledge in a limited natural language

Authors: N.G. Volchenkov

NRNU «MEPhI», Moscow, Russia

ORCID: 0000-0002-1474-6853, NGVolchenkov@mephi.ru

 

Abstract

The prototype of the system for image analysis of geometric bodies, using the interface of logical and visual programming, was developed by the author for experiments with an intelligent robot that is equipped with a video camera. It is assumed that the robot is designed to plan its actions - capture and transfer of bodies. The class of bodies considered by the author is colored polyhedra in local colors. The drawback of the results obtained in the author's previous publication is the absence of an analysis of the relative location of the identified set of polyhedra. Obvious are the difficulties that can arise when trying to automatically (without the participation of a person) to identify the mutual arrangement of bodies. The task is facilitated by the inclusion in the system of the block of interactive training of the robot by man. This person is an operator - a person who can formulate his description of a specific image in a limited natural language. The form of this description is the so-called surface structures of natural language phrases. In this article, the author presents the program of syntactic analysis of surface structures necessary for this purpose. This program is implemented on Prolog - the language of logical programming. To illustrate the results of the training block implemented on Prolog, the author, like in his previous publication, offers a visual programming interface (Visual Basic language) and logical programming (Prolog language). The article presents an example of a concrete image, on which 5 bodies of different colors are revealed. This example allowed us to demonstrate the most typical cases of the relative positioning of bodies, their description on the surface structures proposed by the author and the syntactic analysis of the phrases of this language. An important side effect of syntactic analysis is also presented - the construction of deep structures of a limited natural language. These structures are represented in the form of structures of the language Prolog - on the developed by the author language of deep structures. This view can later be used by the robot directly to plan its actions.

 

Keywords: logical programming (LP); language Prolog; visual programming; the Visual Basic language; database of Prolog; limited natural language (LNL); the surface structure of the phrase on the LNL; learning as introducing new knowledge into the Prolog database; syntactic analysis of the LNL; the language of deep structures (DSL); the definite clause grammar (DCG) – the mechanism in Prolog.