Road Safety and Traffic Control Signs

Road safety remains an important concern in Zambia, where the condition and visibility of road signs and road markings play a significant role in guiding road users. However, assessing these traffic controls through manual inspection can be time-consuming, particularly when analysing large amounts of road footage. This project aimed to develop a web-based system for analysing road signs and road markings from recorded road video footage using computer vision techniques. The specific objectives were to extract key frames from road videos, detect and classify visually identifiable road signs and markings, and develop a web interface for presenting the analysis results.

The system was developed using the Rapid Application Development (RAD) model, which supported iterative prototyping, testing, and refinement. Python and OpenCV were used for video processing and key-frame extraction, while the YOLOv8s object detection model was integrated to support traffic sign detection. Flask was used to develop the web application, with HTML, CSS, and JavaScript providing the user interface and SQLite supporting data storage. The system enabled users with appropriate access rights to upload and manage road videos, process footage, and view available analysis results through the web interface. Testing using Nationalist Road video footage demonstrated that the system could extract 44 frames at five-second intervals. However, no qualifying stop-sign detections were recorded during that test. This result demonstrated the operation of the frame-extraction process but also highlighted limitations in the detection process and the need for further testing with additional videos and traffic sign classes. The project established a foundation for computer-assisted road infrastructure assessment by providing a structured approach to processing road footage and presenting analysis results. Nevertheless, the system was a prototype and did not replace physical road inspections or establish the actual condition of all traffic controls across Zambia. Further development and validation were required to improve detection accuracy, expand the range of detectable road signs and markings, and support analysis of larger video datasets.