Automatic Detection of Bus-Stops and Bus-Crowdedness Using Crowdsourced Data

Abstract

An Advanced urban public transportation system (APTS) uses information and communication technology based solutions to improve the commuters’ experience during their trips using public transportation buses. Generally, the deployment of APTS solutions requires the manual route feedings which is a tedious task specifically when the route changes are frequent. We propose an automatic bus-stop detection solution using the DBSCAN algorithm and the crowdsourced data of commuters. The DBSCAN algorithm finds all the locations where the bus stops during its trip (including the non-bus-stop locations). The bus-boarding event detection of commuters filters out all the bus-stops on the route accurately. The existing solutions for crowdedness detection of the bus use infrared sensors, CCTV camera, treadle mat, and vehicle weighing device. We propose a novel solution that uses the crowdsourced data collected from commuters. The solution is based on processing accelerometer data of commuters in the bus and classifying them as standing or sitting during the course of their journey. The problem is challenging due to the noisy environment inside the bus. The proposed solution classifies the commuters as standing or sitting with 89.72% accuracy enabling the cost-effective crowdedness detection.