Using of behavioral information for enhancing Conditional Random Field-based map matching

Abstract

In this paper we propose an enhancement to our previous Conditional Random Field (CRF) based map matching algorithm in order to make the map matched trajectory smoother and more feasible. The existing algorithm uses one feature, which is the distance with the input coordinate, and has the problem of non-smooth output trajectory. In this work we propose adding a new semantic layer to the map model that depends on the behavioral areas of walking, and to use that information in the map matching algorithm to enhance the smoothness of the output trajectory. The common walking lines of pedestrians will be determined and this behavioral information will be added as a feature to the CRF algorithm. A test version of the new algorithm is applied on a few examples and the results show smoother and more accurate map matched trajectories.