Automated detection and vectorization of road marking

Abstract

This paper presents an innovative method for automatic detection of specific road infrastructure elements and their spatial localization. The algorithm is based on the analysis of the colour spectrum of the objects on orthophotomaps. Source images can be obtained from aerial photogrammetry or terrestrial laser scanning, while digital elevation models can also be included in the processing. The method is based on the identification of the colour composition of road features, their clustering and mathematical transformation into vector form. Subsequently, vectorisation is performed on selected parts of the image, which is the main difference from traditional methods that work with the whole raster image. This approach allows for efficient and automated identification and analysis of features such as road edges, traffic signs or other road infrastructure elements. The resulting outputs can be easily integrated into more complex spatial models of roads and their surroundings. Automating this process brings significant time and cost savings in the creation of digital models of the environment.