Big Data Enabled Vehicle Collision Detection Using Linear Discriminant Analysis

Abstract

Potential lack of witness and monitoring equipments make timely accident rescue a challenge after vehicle collision happens. It is necessary to deploy onboard sensors to achieve the real-time detection for vehicle collisions. However, how to accurately detect collisions by filtering fluctuate data from sensors near vehicle engine remains a problem. In this paper, we take the first step to conduct an experiment and gather the labeled data of 2,700 collisions. Then, a vehicle collision detecting scheme based on linear discriminant analysis (VCD-LDA) in big data enabled vehicular network is proposed. The experimental results show that VCD-LDA scheme discern 99.34% collisions with higher accuracy, compared with other classic algorithm of machine learning.