Využití metod umělé inteligence v řídicích systémech průmyslových podniků

Abstract

The dissertation deals with the processing of images obtained from the industrial environment. A method for filtering tracks after cutting in an image, a newly created binarization method based on pixel sum derivatives, and methods for automatic selection of a binarization method based on image parameters and using artificial intelligence methods have been described. Methods for finding the positions of characters in the image with subsequent character recognition were implemented. Three different methods were implemented for character recognition - OCR, Hopfield neural network and neural network with error backpropagation algorithm. Part of the work is a comparison of the resulting success of individual methods for image recognition on images obtained by the commonly used image processing process and on a method that uses all implemented methods to increase image quality and success of character recognition, which are described in the work.