PROBLEMS OF ORDERING OBJECTS ON IMAGE BY USING NEURAL NETWORKS AND HEURISTIC ALGORITHMS

Abstract

Currently, in various branches, such as space and automotive industry, computer vision algorithms are widely applied. In different tasks of computer vision, it may be necessary to determine the order of the objects in space. At first glance, this task may seem insignificant, but the quality of the images strongly depends on shooting conditions. There is a set of solutions for finding the order of objects on images. At the same time, most of the solutions are not very useful for working with human-made objects. One of the special cases of this task is the ordering objects according to a flat table. The main problems in solving this task are photographs with perspective distortions and small displacements of objects relative to each other. The article presents the results of the research conducted in this area. Heuristic and neural network approaches to the solution are considered. The neural network method showed the most sufficient results. The authors named the design of described neural network as neurotabulator.