generate the time-stamped Sequence of Visited Locations (SVLs) of the mobile objects. The sequence is constructed by labeling each of the GPS observations of the trajectory using the ID of their intersecting Geometries of Interest (GOIs). In this paper, we enhance the performance of the state-of-the-art scheme for constructing the GOIs of a mobile object by proposing a data aggregation and outlier detection method. Our experimental results using geometric similarity metrics show that our improved GOI construction method outperforms the baseline methods by constructing the GOIs remarkably more geometrically similar to the real world GOIs. The geometric similarity metrics are only applicable when we have access to the geometries of the real world GOIs (ground truth). To be able to analyse the performance of the GOI extraction methods in environments which we do not have access to the ground truth, we propose two useful spatio-temporal metrics to measure the quality of GOIs based on the quality of the generated SVLs based on them. Our experimental results show that these two metrics are able to discriminate between the results of our different outlier detection methods and select the best scheme without using any external knowledge about the geometries of the real world GOIs.

Keywords