Using descriptors for image recognition
Abstract
Methods for finding singular points and forming their descriptors are considered. The aim is to study the existing search methods and determine the feature point descriptors to select the best match between feature point detectors and their descriptors for different types of images. In this paper, we carry out a comparative analysis of the PCA-SIFT, ORB, BRISK, AKAZE methods, which detect singular points and describe their descriptors in the image. An algorithm has been developed that, based on the work of these methods, groups photographs according to the degree of similarity. A comparative analysis of the methods is carried out on several collections of photographs of different quality at different compression rates and at different values of the descriptor matching parameter.