Key points are not available for this paper at this time.
Car accidents result in numerous deaths and disabilities daily, a significant number of which are due to delayed treatment and subsequent accidents.Automatic car accident detection can reduce the response time for rescue teams and nearby vehicles, thereby enhancing rescue efficiency and traffic safety.In this paper, we introduce an automatic car accident detection method utilizing Cooperative Vehicle Infrastructure Systems (CVIS) and machine vision.Firstly, we establish a novel image dataset, CAD-CVIS, to boost the accuracy of accident detection using intelligent roadside devices in CVIS.The CAD-CVIS dataset features a variety of accident types, weather conditions, and locations, which enhances the adaptability of detection methods across different traffic scenarios.Secondly, we develop a deep neural network model, YOLO-CA, based on CAD-CVIS and deep learning algorithms for accident detection.This model incorporates Multi-Scale Feature Fusion (MSFF) and a dynamicweight loss function to improve the detection performance of small objects.
A Thu, study studied this question.