Önvezető járművek HD-térképeinek adatminőségi és topológiai vizsgálat

Abstract

The thesis aimed to investigate the data quality and topology of high-definition (HD) maps, which are essential for the operation of self-driving vehicles. In the first part of the research, I analyzed the formats of HD maps and their characteristics. Then I adapted a data quality model known from the literature to the maps’ specific structure and application environment. In this context, the dimensions of accuracy and consistency were reinterpreted, with particular reference to their spatial and topological extension. Studies on the measurability of data quality have provided a basis for comparing different map formats and have highlighted the need for structured, scalable quality assessment. In addition, I have developed an implementation of an OpenDRIVE model in a relational database and a software tool for loading it.

In the next part of the thesis, I investigated the possibilities of map information retrieval from in-vehicle camera images, in particular, the applicability of processing methods based on community data collection and artificial intelligence. Based on experiments with the Mapillary platform, I demonstrated that traffic signs and pavement markings can be recognized with high accuracy and organized into a structured database, making the method suitable for partial updates of HD maps. However, the geometric and consistency errors detected during the validation of the system highlighted the limitations of automated processing and the need for standardization.

In the final phase of the research, I developed a proprietary graph-based topological formalism and a set of verification rules that can test the consistency of the road network structure of HD maps. The method was implemented in the MATLAB environment and validated on synthetic and real data, including OpenStreetMap and OpenDRIVE samples. The results demonstrated that formal and structural errors in topological relationships can be automatically identified, and improvement options can be defined based on the graph structure. The thesis thus contributes to improving the reliability of map databases and provides a basis for future developments in the field of automated map processing and validation.