Published on December 2021 | Artificial Intelligence, Embedded System, Camera Design

CV based accident avoidance system through video surveillance using Deep Learning Algorithms.

Detecting accidents using computer vision become useful but time-consuming task. Here we proposed a neoteric framework for detecting road accidents. In the proposed framework for surveillance footage, a Mask R-CNN object detection algorithm is used, followed by an efficient tracking algorithm. Anomalies in the vehicle's speed and trajectory are used to predict the likelihood of an accident occurring after an accident involving another vehicle. Using the proposed framework in conjunction with CCTV footage, it is possible to achieve very high Detection Rate. The framework was tested under a variety of conditions. This included bright sunlight and low visibility, as well as precipitation such as hail and snow. According to the findings, an effective framework has been established that will allow for the development of real-time vehicular accident detection algorithms.