Methods for Identification and Vectorization of Traffic Elements from Orthographic Images

Image documentation, particularly orthographic images, is a common and important foundation in traffic engineering, primarily used for mapping the evaluated location and displaying its spatial layout. The use of these images is also suitable as a base layer in graphic software. When subsequent work involves maintaining the current state or performing only partial reconstructions of the traffic space, the creator is required to reconstruct the existing condition. Design proposals in the form of studies are then based on reconstructing the traffic area from the current image. On raster images, manual tracing of elements such as road boundaries, horizontal traffic markings, or other features is typically performed, which are then vectorized into curves. If the area is complex within its existing layout, the designer often spends excessive time on vectorizing selected elements within the image. The aim of this article is to present methods for automating and processing image data that enable fast and efficient vectorization of selected traffic infrastructure elements. The intention was to develop a universal solution that does not rely on sophisticated machine learning systems, but instead emphasizes a step-by-step processing pipeline that is fully transparent and can be precisely controlled or customized by the user. As part of the study, the proposed automated method was compared with manual vectorization in terms of time efficiency. The results showed that automation can accelerate the process by approximately more than 2 times.