Loading [MathJax]/extensions/MathMenu.js
Accident Detection Using Convolutional Neural Networks | IEEE Conference Publication | IEEE Xplore

Accident Detection Using Convolutional Neural Networks


Abstract:

Accidents have been a major cause of deaths in India. More than 80% of accident-related deaths occur not due to the accident itself but the lack of timely help reaching t...Show More

Abstract:

Accidents have been a major cause of deaths in India. More than 80% of accident-related deaths occur not due to the accident itself but the lack of timely help reaching the accident victims. In highways where the traffic is really light and fast-paced an accident victim could be left unattended for a long time. The intent is to create a system which would detect an accident based on the live feed of video from a CCTV camera installed on a highway. The idea is to take each frame of a video and run it through a deep learning convolution neural network model which has been trained to classify frames of a video into accident or non-accident. Convolutional Neural Networks has proven to be a fast and accurate approach to classify images. CNN based image classifiers have given accuracy's of more than 95% for comparatively smaller datasets and require less preprocessing as compared to other image classifying algorithms.
Date of Conference: 01-02 March 2019
Date Added to IEEE Xplore: 29 August 2019
ISBN Information:
Conference Location: Bangalore, India
References is not available for this document.

I. Introduction

Over 1.3 million deaths happen each year from road accidents, with a further of about 25 to 65 million people suffering from mild injuries as a result of road accidents. In a survey conducted by the World Health Organisation (WHO)on road accidents based on the income status of the country, it is seen that low and middle-income or developing countries have the highest number of road accident related deaths. Developing countries have road accident death rate of about 23.5 per 100,000 population, which is much higher when compared to the 11.3 per 100,000 population for high-income or developed countries [1]. Over 90% of road traffic related deaths happen in developing countries, even though these countries have only half of the world's vehicles. In India, a reported 13 people are killed every hour as victims to road accidents across the country. However, the real case scenario could be much worse as many accident cases are left unreported. With the present data, India is on the way to the number one country in deaths from road accidents due to the poor average record of 13 deaths every hour, which is about 140,000 per year [2]. An accident usually has three phases in which a victim can be found.

Select All
1.
"Global status report on road safety 2015", World Health Organization, 2019, [online] Available: http://www.who.int/violence_injury_prevention/road_safety_status/2015/en/.
2.
S. Prabakar et al., "An enhanced accident detection and victim status indicating system: Prototype", India Conference (INDICON) 2012 Annual IEEE, 2012.
3.
"Lexus Enform", Lexus, 2019, [online] Available: https://www.1exus.com/enform.
4.
"OnStar Safety and Security Services", Onstar.com, 2019, [online] Available: https://www.onstar.com/us/en/services/safety-security/.
5.
"SOSmart automatic car crash detection and notification app", SOSmart automatic car crash detection app, 2019, [online] Available: http://www.sosmartapp.com.
6.
C. Kockan, Communication between vehicles, 2008.
7.
Zeng Yuanyuan, Deshi Li and Athanasios V. Vasilakos, "Opportunistic fleets for road event detection in vehicular sensor networks", Wireless Networks, vol. 22.2, pp. 503-521, 2016.
8.
Christian Szegedy et al., "Rethinking the inception architecture for computer vision", Proceedings of the IEEE conference on computer vision and pattern recognition, 2016.
9.
Christian Szegedy et al., "Going deeper with convolutions", Proceedings of the IEEE conference on computer vision and pattern recognition, 2015.

Contact IEEE to Subscribe

References

References is not available for this document.