OBJECT CLUSTERIZATION METHOD IN PICTURES BASED ON FEATURE SELECTION

Abstract

The article describes the development of a method that allows you to create clusters based on selecting feature features. In todayʼs world, the entertainment industry on the Internet is developing rapidly, creating a demand for better products. This factor has led to the use of artificial intelligence not only in science but also in entertainment. Currently, applications that allow you to create animations of objects in photos are gaining popularity. This article presents an approach to solving the problem of defining objects for animation. To classify and further identify objects, their characteristics should be determined. This is one of the options for abstraction, in which the input set of properties of the object is reduced to the minimum required number of features by which you can identify the object.

The algorithm can be used to determine the main features of objects, such as area and perimeter, radii of inscribed and circumscribed circles, sides of the described rectangle, number and relative position of angles, gradient of the object histogram. Based on these features clustering and classification of the image are implemented. The artificial neural network was trained on image samples, each class contained from 2528 to 16185 images of 64x64 pixels. 1000 images of objects of each class were then selected for testing. The success of recognition based on a convolutional neural network was evaluated. According to the results, we can conclude that the smaller the invariance of the class, the greater the accuracy of recognition. The amount of data in the training sample has little effect on the accuracy of the algorithm. After calculating the intensity gradient, you should divide the image into a cell and build a histogram of the gradient object for each pocket of cell; the histogram module corresponds to the intensity gradient at the point.