Rapid urbanization has led to a variety of problems in road traffic, such as recurrent congestion and traffic accidents. To mitigate these problems, it is necessary to monitor and analyze traffic volumes in order to anticipate or detect anomalies in road traffic. This work proposes a solution for traffic accident scene recognition using computer vision applied to drone imagery. A Mavic Air 2S drone collects geo-referenced aerial images of the portion of road to be monitored, and the data is annotated, trained and tested on Roboflow. The results of the experiment give an overall object detection rate of 95.3%, accuracy of 94.4% and recall of 91.3%. Geographical coordinates of images containing accident scenes are extracted and stored in a database for statistical and rescue purposes.
Real-Time Road Accident Scene Recognition Using Computer Vision Applied to Drone Imagery
2024-11-08
724951 byte
Conference paper
Electronic Resource
English
Advanced computer vision for extracting georeferenced vehicle trajectories from drone imagery
ArXiv | 2024
|Real-time road scene classification using infrared images
Tema Archive | 2010
|