Mining Traffic Condition from Trajectories

Advances in wireless transmission and increasing quantity of GPS and other positioning sensors carried by vehicles flood us with amount of moving objects data. Trajectories are mixture result of moving entity´s purpose and environment constraints. The large amount of trajectory implies considerable quantities interesting road condition. In this paper, we build constraints matrix to express environment constraints, employ road linear reference system and road segmentation to preprocess raw trajectories data, and propose data selecting based on entity similarity and spatio-temporal similarity. The trajectory mining is classified as road condition finding near intersection and road information away from road intersection. The former focuses on finding road intersection turn information, while the latter focuses on extracting live route condition. The experiment results show that the trajectory mining algorithm is effective and efficient to discover traffic conditions.