Traffic surveillance footage requires constant monitoring in order to detect deadly accidents and to react appropriately. However, it is tiresome and mistake prone to regularly supervise them with humans. A sophisticated and quick system for detecting road accidents is urgently needed due to the rise in fatalities from accidents on the roads. After accidents, a few crucial seconds frequently makes a lot of difference. There is a need for a system that can immediately recognize emergency situations and notify the police and hospital about them. This study proposes a real-time traffic incident detection and alarm system that is based on a computer vision technique, with the goal of achieving a penetrating impact in boosting the associated demand for road traffic security and safety. The vehicle detection and tracking model first assigned a distinct serial number (ID) to each vehicle in order to detect and track the movements of the vehicles. This was done using the YOLOv3 object detector and DeepSORT tracker with region of interest interception. The mean average precision (mAP) of this model was 98.2%.
A Novel Approach for Accident Detection and Localization Using Deep Learning
2023-11-23
1479218 byte
Conference paper
Electronic Resource
English
Accident Detection Using Deep Learning
IEEE | 2020
|Acoustic accident detection system through accident sound deep learning
European Patent Office | 2022
|Acoustic accident detection system through accident sound deep learning
European Patent Office | 2022
|