Pedestrian Travel Distance Estimation Using Optical Flow and Smartphone Sensors

Abstract

The extensive proliferation of smartphones presents a new opportunity for precise indoor and outdoor location-based services. Although various smartphone-based positioning technologies have been investigated, no practical and economical method has been developed where global positioning system (GPS) cannot be applied. Since it is difficult to estimate travel distance accurately with a low-cost inertial sensor embedded in a smartphone, several methods have been proposed to estimate stride. However, the estimation of stride is not accurate as it depends on walking posture. In this article, a novel approach is proposed to estimate the travel distance of the smartphone using optical flow (OF) technology, which is applied to the video image captured from the smartphone camera. Several algorithms have been developed to solve various problems that occur when a user walks holding a smartphone. The feature trace method proposed in this article improves the travel distance estimation accuracy greatly. Several motion compensation methods for the OF are proposed also using smartphone sensor data. The performance of the proposed methods is verified by experiments in three different places with two kinds of smartphones. The estimation accuracy of the travel distance by the proposed method is obtained from all the experimental data. As a result, the average travel distance error rate is 2.6%; for the 5% distance error rate, the error CDF is 86%; and for the 10%, it is 100%. The proposed methods could be applied to track the smartphone of the pedestrian with precise position accuracy indoors and outdoors.