A climate-invariant intelligent real-time system to recognise and report road accidents
J Sukanyathara, Lalu Seban · 2017 IEEE Region 10 Symposium (TENSYMP) · 2017
Due to the slow response time to road accidents, most of the victims remain unattended and might be prevented from getting enough care or medical aid in time. Road traffic monitoring systems may also become inefficient when visibility of the scene is hindered by certain unfavourable climatic conditions. The risk for accidents during a heavy rain, foggy weather, or snowfall is higher as well. This paper proposes a climate-invariant system which robustly identify road accidents even in challenging outdoor conditions by analysing road videos captured with a camera placed over a pole. This is an intelligent memory efficient system that works in real time for emergency alert of road accidents. The moving vehicles are segmented and tracked throughout the scene. The proposed algorithm distinguish between the normal moving vehicles and those affected by collision by closely analysing the deceleration pattern of the vehicle and its deviation from the normal motion in real time. Motion cues and trajectories are analysed to accurately identify the variation from normal traffic. The ultimate aim is to enhance the capabilities of intelligent transportation systems to minimise the accident response time and to introduce a system which is adaptive to any climate, accurate, inexpensive both computationally and economically.