A Fast-Mining Method for Target Behavior Pattern Based on Trajectory Data

With the development of location-acquisition techniques, a large amount of trajectory data has been accumulated in the early warning and surveillance system, which contains lots of information and knowledge. By using unsupervised clustering of data mining technology, a new fast mining method for target behavior pattern is proposed in this paper. First, target trajectory data is compressed by using Douglas-Peucker algorithm to reduce the amount of calculation. Then, a new Hausdorff distance with direction recognition is designed to measure the similarity of target trajectory. Finally, an improved Density Peaks Clustering (DPC) and K-medoids algorithm is proposed to mine the regular behavior. In this paper, two datasets of simulation and measurement are used to test the performance of the proposed method. Compared with the existing classical algorithms, the verification results show that the proposed method can detect the target behavior pattern more quickly and accurately in a large number of complex historical trajectory data.