Students of Novosibirsk State Technical University (NETI) are creating an intelligent system for the rapid detection and fixation of defects in the road surface and traffic controls. The work is carried out within the framework of student project activities.
The development of this system represents an important step towards improving road safety and quality, as well as effective management of road infrastructure.
"The system is based on computer vision and is able to identify various defects, such as pits, damage to road signs and traffic lights, highlighting them on the video stream. As part of the project, students collected and marked up a dataset (data set) of more than 6,000 photographs (road pits, broken traffic lights, an erased zebra, a tree blocking a road sign, etc.). The project team has trained a convolutional neural network model for detecting defects on this dataset," said Egor Antonyants, project manager, Assistant Professor at the Department of Automated Control Systems at NSTU-NETI.
According to him, the system will be able to identify various defects on the video stream: holes and cracks, broken road signs, non-working traffic lights, etc. — and transfer their geolocation to the visualization module on the city map. This development can be used in the field of urban improvement to update the state of infrastructure. The system will be useful for drivers, road services, and government agencies.
"We are currently at the stage of continuous testing and improvement of neural network defect recognition in the recorded video stream. In the future, we plan to add the ability to work with the interface and connect an active video stream from a video camera," added Ruslan Gavrilov, one of the main developers of the project, a fourth—year student at the Faculty of Automation and Computer Engineering.
In the future, the software product can be installed on urban transport to fix defects and then transfer the received data to the municipality. The prototype will be ready by September, and the developers plan to present the full version in December this year.
It should be noted that scientists from NSTU-NETI and the Territorial Administration of Highways of the Novosibirsk Region have launched a project to develop software for creating digital models of highways.