Recognition of Damaged Arrow-Road Markings by Visible Light Camera Sensor Based on Convolutional Neural Network

Area of Science:

• Computer Vision
• Artificial Intelligence
• Road Safety Engineering

Background:

• Road markings provide critical information for drivers and advanced driver assistance systems (ADAS).
• Damage to road markings, such as arrow-road markers, compromises safety and can lead to accidents.
• Automated identification of damaged road markings is an under-researched area.

Purpose of the Study:

• To develop and evaluate a method for automated recognition of damaged arrow-road markings.
• To improve the robustness of road marking identification systems for ADAS applications.
• To address the safety risks associated with degraded road marking visibility.

Main Methods:

• A convolutional neural network (CNN) model was employed for image recognition.
• The system utilizes visible light camera sensor data.
• The method was tested on six diverse road marking datasets, including KITTI and Málaga urban datasets.

Main Results:

• The proposed CNN-based method demonstrated superior performance compared to conventional approaches.
• The system effectively recognized six types of arrow-road markings, even when damaged.
• Experimental results validated the method's accuracy across multiple datasets.

Conclusions:

• The developed CNN method offers a reliable solution for identifying damaged road arrow markings.
• This technology can significantly enhance the safety and functionality of ADAS.
• Further research in automated road marking recognition is crucial for future transportation systems.