| Abstract [eng] |
The goal of this paper is to research feature extraction and classification methods possibilities for numerals recognition. The state of the art of the number recognition methods is presented in the first chapter of the thesis. In the second chapter, some background mathematical skills that will be required to understand the process of methods is given. There are a lot of feature extraction methods for image processing. Presented paper deals with two feature extraction methods: Principal Component Analysis (PCA) and Histogram of Oriented Gradients (HOG). Various classification methods can be used for features classification. In this paper, Artificial Neural Networks (ANN), Support Vector Machines (SVM) and Random Forests (RF) have been applied to numerals recognition. On the third section of this paper, algorithms of feature extraction and classification methods are shown, which are written with C# programming language in Visual Studio 2015, using OpenCV library. The implemented algorithms for numerals of LED segments recognition are investigated and results are analyzed. All methods, which are mentioned, were trained and tested on a data set of 1000 different images. Experimental results show that the recognition accuracy is the best (99,4 %), when we use HOG with SVM. |
