Розробка алгоритмічного та програмного забезпечення методу випадкових лісів в задачах класифікації зображень

Abstract

In this thesis work the problem of increasing the reliability of the classification of noisy images. A comparative analysis of methods for image classification and proposed approach to solving this problem in the form of a new method developed by the image classification, which, unlike the existing uses of the forest and the occasional local binary patterns as descriptors of the image. On the basis of this method created a software application designed for image classification using random forests. To study the effectiveness of this method used in the base color images Aerial Image Classification. Experimental research of a method carried out by using a medium Matlab and R.