26 October 2021

Together with Moscow State University, Avtodor Creates Artificial Intelligence to Detect Defects on Sound Barriers

An artificial intelligence-based solution has been developed and tested to automatically detect and digitize defects on sound barriers and to monitor the road infrastructure condition.

As part of the pilot project, the Russian Highways SC and researchers of the STI Competence Center for Big Data at the Lomonosov MSU with participation of OZMK LLC have created artificial intelligence to control condition of the road infrastructure and detect defects on sound barriers. This solution will help in monitoring condition of the road infrastructure facilities for timely upgrade and control of warranty obligations.

Works started based on results of evaluation of various barrier designs available at the test ground created by the State Company last May at stage 1 of the future M-12 Moscow - Nizhny Novgorod - Kazan highway.

Artificial intelligence is capable to detect corrosion, dints, graffiti, pollution and other damages on the sound barriers. The solution is also capable to keep track of the number of sound barriers and analyze their color.

Damages are checked for conformity with warranty requirements of STO AVTODOR 2.9 and international standards. Detected damages are typical of barriers used on motor-roads for any technical category.

The operation algorithm of the solution suggests that each car should be equipped with a device containing a camera, a GPS module, a microcomputer and an illumination sensor. The latter allows for the algorithm adaptation to weather conditions, enabling accurate registration of defects both in sunny and cloudy weather. When a car passes by sound barriers the device continuously records video, geo-data and illumination level. The acquired data is analyzed by the artificial neural network which detects defects on sound barriers. Based on geo-data, the web interface builds a map where each detected defect is marked with a dot and accompanied by a relevant image and information on the damage area.

This solution will allow for monitoring the condition of sound barriers made of various materials, violation of warranty features, as well as for efficient upgrade planning and implementation. There is also possibility to adapt the technology to various barrier designs and materials with account of the damage specifics.

In addition to sound barriers, the developed solution can keep track and evaluate the condition of other road infrastructure facilities. These include road signs, traffic light, lamps, solar panels, crosswalk and other road marking (including temporary marking), etc. The solution also makes it possible to automatically detect and determine cracks and other damages of the roadbed.

Future plans of the researchers envisage extension of functionality and additional model learning, including based on the data acquired in the road monitoring process.