Abstract:
The paper presents methods for highlighting the contours of objects on the image, two methods of recognizing the images of geometric shapes are selected for a detailed study. The first method is based on the selection of the contour and counting the number of vertices. The second method uses a modern approach using artificial neural networks. The results of comparing the recognition accuracy of these methods on the same test samples are presented; the accuracy metric is chosen as the quality criterion for the developed model. This metric is an indicator that describes the overall accuracy of the model for all classes.
Keywords:artificial neural networks, pattern recognition of geometric shapes, contour analysis, Suzuki and Abe algorithm, multilayer perceptron.
Fulltext:
PDF file (500 kB)