| Abstract [eng] |
The aim of this thesis is to investigate the feasibility of micromobility vehicle classification using magnetic field sensors. The main part of the research focuses on classifying experimentally collected magnetic field signatures using machine learning models. The first part of the thesis analyzes scientific literature. Methods applicable to the estimation of micromobility vehicle parameters are reviewed, and the advantages and disadvantages of different machine learning models for classification are presented. The second part describes computer simulations using the finite element method. The necessary requirements for micromobility vehicle parameter estimation systems – installation depth, array spacing and sampling frequency – are established. The influence of magnetic inclination on the recorded signatures of micromobility vehicles is discussed. The third section presents the methodology for micromobility vehicle classification. Methods for data collection and preparation, signal processing, feature extraction, and the training and validation of machine learning models are described. The fourth section analyzes the results obtained from applying machine learning models to the classification of micromobility vehicles. Data analysis is used to support the selection of methodological steps and their parameters. General classification results are presented and compared using statistical methods, and the performance of the highest-scoring model is analyzed in detail. |
