期刊名称：International Journal of Computer Trends and Technology
电子版ISSN：2231-2803
出版年度：2019
卷号：67
期号：6
页码：31-36
DOI：10.14445/22312803/IJCTT-V67I6P104
出版社：Seventh Sense Research Group
摘要：The quick growth in the number of mobile devices such as smart phones, wearable devices, tablets, sensor enabled vehicles etc. with a large array of sensors like GPS, Accelerometer, Gyroscope, Compass, Magnetometer, Camera etc. enables a new sensing paradigm known as Crowd Sensing. Traffic anomalies can occur due to various events like accidents, functions, celebrations, protests, disasters etc. In this paper we propose an architecture that employs crowd sensing to detect traffic anomalies and uses social media data to determine the authenticity of identified anomalies. Our prototype architecture includes an Android based navigation application for the client and a combination of J2EE application server and Hadoop as the backend. The client application consists of an interface to report traffic anomalies apart from the basic navigation features. Anyone using this app can report an anomaly that he encounters in his route. Whenever a user reports an incident, a tweet with the exact location and incident details are posted automatically to the twitter account managed by the application. Using Recursive EM Algorithm, the authenticity of the reported anomaly is verified and if it is genuine, all the users in that particular route will get notified in advance. The system will also suggest the best possible alternative route to the same destination. The system also provides a web interface for the traffic authorities to monitor the anomalies in their locality on a realtime basis and can respond to it very immediately. Hadoop based infrastructure which is deployed in the backend is able to process massive GPS data collected from the users using MapReduce framework. The system has been tested successfully in a simulated environment using Android emulator and GPS Location Spoof application.
关键词：Traffic Anomaly Detection; Truth Estimation; MapReduce; Crowd Sensing; Map-Matching
