Road markings are essential for driving. Their characteristics are surveyed with dynamic retroreflectometers and are provided in an aggregated form to managers, usually averaged on 100 meter intervals. We have developed and tested a road marking analysis method that represents the measured signal as images and allows renderings at finer scales, such as the marking segment scale. To do this, we used several artificial intelligence methods. First, an architecture derived from UNet was proposed for the automatic segmentation of marking elements. To check the consistency of the maps produced, we tested several techniques used in natural language processing with architectures such as LSTM and an attention mechanism derived from Transformers. Since there exist no annotated databases for this type of data, a Model-View-Controller software solution was developed to visualize, annotate and enhance the data. This led to the implementation of a processing chain as well as incremental learning methods. This paper presents both the software for processing data and visualizing the results, the methodology implemented and the first results obtained.