Analýza prechodov pre chodcov pomocou obrazov

Abstract

The aim of this diploma thesis is to find out the possibilities of image processing for the purpose of crosswalk detection in order to detect pedestrian crossings and to compare possible solutions. This diploma thesis is focused on processing pedestrian crossings from the perspective of autonomous vehicles. To create a functional detector, it is advisable to experimentally test several approaches of image processing. The main differences in the proposed detectors are in data preprocessing. In a training approach, it is necessary to use data that must be extracted from available materials. In connection with the processing of pedestrian crossings using images, this thesis compares existing solutions, identifying the possibilities of image processing and design along with the implementation of the detector using the found libraries. The main parts for creating a detector are extracting the data for training, data pre-processing and subsequent use in creating the final detector. A pedestrian safety detector is important in the autonomous driving industry.