The RetroTek team in Ireland are working on a project ‘AI Classifying Road Markings‘ using the Intel Movidius Neural Compute Stick technology which allows the RetroTek technology to utilize and deploy deep neural network and computer vision applications. Thus, providing cutting edge solutions for deploying deep learning and computer vision algorithms on the RetroTek device at ultra-low power. This project will enable the RetroTek Road Marking Retroreflectometer Technology to Identify and Classify the Quality of Pavement Road Marking using Computational Neural Networks (CNNs).
The RetroTek AI-assisted technology enables the classification of the quality of road markings day & night which was mentioned as a key requirement in the CEDR Premium Project for the Conference of European Directors of Roads (CEDR) to assist in maintaining safe roads i.e.
- Identification & classification of the type /shape of the pavement marking.
- Quantify the paint/thermoplastic wear and coverage patterns of these markings (lines, arrows, symbols and text messages) that have been damaged by traffic & weather. Good quality visual coverage is required for day time vehicle operation.
- Quantify Measurement of the amount of retroreflective bead material coverage that on the markings which is essential for marking night visibility and safe road use at night.
Road markings, when heavily damaged or worn away, cannot be classified using conventional machine vision techniques. This development will assist road pavement authorities and maintenance contractors to assess the day and night visibility performance of pavement marking conditions and to predict and plan pavement marking maintenance where specifically required.