Detekce chodců pomocí příznakového rozpoznání

Abstract

This thesis is focused on creating pedestrians detectors in image using pattern recognition renowned features. At first are introduced the basic principles of feature-based methods, which are histograms of oriented gradients, Haar-like features and method of Local Binary Patterns. The next part are introduced classifiers Support Vector Machines and Adaptive Boost. The result of this thesis is the implementation these methods using the OpenCV library and experimentally verify and compare their functionality.