Key points are not available for this paper at this time.
Particular methods exist to compare images based on comparing knowledge about images as a whole; however, for reliable operation of such algorithms, the image should have significant brightness jumps in most areas, heterogeneous scene details, and minimal distortions caused by affine transformations.This study aims to improve the efficiency of video information computing systems of machine vision in data processing by using methods and organization algorithms that allow real-time estimation of the moving object's location and 3D identification.For that, this paper solves a set of base theoretical problems, combining the choice of hardware for obtaining and processing information, determining the coordinate origin, defining the reference plane of the underlying surface, dividing images into levels to achieve higher processing speed, and determining the spatial coordinates of image points from the stereo system.This paper reviews existing image acquisition systems and considers correlation functions for the 3D identification problem.In developing an algorithm that performs real-time 3D identification, the problem at hand is formalized, and the number of levels of the image pyramid is selected, considering the problems arising in 3D identification.
Alexandrov et al. (Wed,) studied this question.