Development of invariant features of nonconvex images for system of automatic recognition of three-dimensional objects

The ubiquity increase of production volumes leads to the introduction of automatic conveyors and assembly lines. An integral part of it is the robotic units, equipped with technical vision. System of automatic recognition is used therein as a machine vision. The main part of any SAR is a vector of image features. During the past five years, various of researchers have published many algorithms for calculating object features that allows to recognize three-dimensional objects [1, 2, 3 and 4]. This article describes an algorithm for calculating the dimensionless features of flat binary nonconvex images that are invariant to a shift, rotate, and scale change of the image of the object in the scene. The basis of developed features is a vector of diagonal features of form (created by author) of images of projections of three-dimensional objects, that describes the shape of the convex projection of the object. Selecting the basis for the development of new features is determined by the fact that using the basis of diagonal features of form, image classification model was created. It allows to use this features for solving the problem of recognition of randomly arranged three-dimensional objects by two images.