Creation of Spatial Structured Trajectory Based on Stop and Move Model in Order to Semantic Enrichment

Nowadays, with considerable developments in technology, the accessibility and usage of positioning devices has been increased. These systems facilitate generation of position data as streams of spatio-temporal data, so-called trajectory. In recent years, research related to trajectory data management has mostly focused on the techniques of storage/retrieval, modeling, and data mining and knowledge discovery from trajectory data. These studies mainly emphasize on geometric aspects. However, emerge of different applications area from shipment tracking to geo-social networks generate motifs to detect semantic behavior (pattern) of moving objects. Semantic patterns include not only geometric patterns, but also the knowledge derived from the data relating to the geographical and application domains. Most of the studies in the field of semantic trajectory are based on offering different ways to add meaning to trajectories and little work has been carried out on preparing data for the semantic enrichment. Actually, process of knowledge discovery and semantic enrichment is a computing framework that commonly involves several steps.
In this paper, an effective method as a multi-step computing framework is proposed and the results have been evaluated. In the proposed methodology, the outputs of levels are raw cleared data, spatio-temporal trajectory, and structured trajectory, respectively. At the first level, which includes techniques for cleaning raw data collected by moving objects, the algorithms for data cleaning from outliers are assisted and removal of unnecessary data are provided. To identify and eliminate the outliers from data set, a two-step algorithm is provided based on three sigma rule. Another step that needs to be done is the use of compression techniques for detection and removal of additional data. Compression techniques are typically based on linear simplification and use distance functions to approximate data. However, these methods are not able to keep stop and move points of trajectories. In this study, a compression method based on velocity of points is provided. The proposed method is based on combination of two functions. The first function is based on instant velocity of points to calculate distance function. Another function is perpendicular Euclidean distance. However, in this function instead of constant velocity, assumption of constant acceleration has been used in the interpolation. The results of implementation show that the suggested algorithm is able to reduce the number of points at the same time keep important point in trajectories. At the second level of abstraction, spatio-temporal trajectories have been derived from cleaned data. In this stage, trajectory identification is based on type of data and application. In this work, based on the specified application, which is transportation and traffic management, daily trajectories were identified from cleaned data. The final step in preparation of data for semantic enrichment is producing structured trajectories as stop and move episodes. For this task, a method based on velocity of points is implemented. In this method, the moving object data and environment in which it moves is used to identify episodes. In this paper, the proposed algorithm in the semantic enrichment framework has been implemented on a real trajectory data and evaluation results show the effectiveness of the method.