Publication Type:

Conference Paper

Source:

Advances in Distributed Computing and Machine Learning, Springer Singapore, Singapore (2021)

ISBN:

9789811542183

URL:

https://link.springer.com/chapter/10.1007%2F978-981-15-4218-3_25

Abstract:

Wide variety of road types like intersections, highways poses a real challenge to the computer vision algorithms. Hence, there is a need of efficient algorithm to detect the accident on road and also evaluate the severity of the incident. This can be used to improve the emergency services response time. The work demonstrated in this paper aims to develop such an algorithm by modifying existing CCTV surveillance system. In this work, the accident is detected by the dispersion in the motion field of the vehicles during collision. Motion field of the road is obtained from the optical flow of the video frames. The moving objects in the frames are segmented and tracked. The dispersion in the angle vector of the optical flow is derived for each of the moving object. The dispersion of angle vector for each object is monitored, and deviation of the same from the threshold is determined as an accident. The harshness of the accident can be found by the range of dispersion of the motion field. The algorithm developed here is capable of detecting accidents between any types of moving objects.

Cite this Research Publication

S. Veni, Anand, R., and Santosh, B., “Road Accident Detection and Severity Determination from CCTV Surveillance”, in Advances in Distributed Computing and Machine Learning, Singapore, 2021.