Detekce chodců na embedded systémech

Abstract

Pedestrian detection has attracted great attention in recent years, especially in safety applications. For example, it can be used in security cameras in public places or in automotive systems, which can prevent potential accidents. For detection, embedded systems such as Raspberry PI can be used. These are small and affordable computers that can be placed almost anywhere. The goal of the thesis is to test and attempt to optimize the recognition techniques for such devices. Pedestrian detection in images processing is solved, for example, by Histograms of Oriented Gradients, abbreviated as HOG. Optimization of the algorithm is ensures by background subtraction, which greatly accelerates detection process. These algorithms are throughly described in this thesis. Further pedestrian detection techniques will be described her. Part of the thesis is also a comparison of detector rate and speed on selected devices.