Accident Detection through CCTV Surveillance

Traffic has become an integral part of people's life in recent years, influencing a wide range of human activities and services. Accidents have been a leading cause of death worldwide, particularly in India. Approximately more than 80% of accident related deaths cases are caused by lack of timely assistance to accident victims rather than the accident itself. One of the basic reason for the vehicle accident is the vehicle speed and there are many others too. Many lives can be saved if there was a proper emergency service that could get accident information in time. Computer vision-based accident detection on traffic video surveillance is a advantageous but also a challenging task. This paper proposes an automated accident detection from traffic videos. Here the frames are extracted from the video shots. Key-frames are identified using the histogram difference, later features are extracted from it using the VGGNET and finally classified as accident or non-accident. The proposed methodology has been compared with the literature methods and found out to be the best.