Local Behavior Analysis for Trajectory Classification Using Graph Embedding

Abstract

Understanding motion patterns is of great importance to analyze the behavior of objects in the vigilance area. Grouping the motion patterns into clusters in such a way that similar motion patterns lie in same cluster and the inter-cluster variance is maximized. Variation in the duration of trajectory patterns in terms of time or number of points in them (even in the trajectories from same cluster) make it more difficult to correctly classify in respective clusters as a bijective mapping is not possible in such cases. In this paper, we have formulated the trajectory classification problem into graph based similarity problem using Douglas-Peucker (DP) algorithm and complete bipartite graphs. Local behavior of objects has been analyzed using their motion segments and Dynamic Time Warping (DTW) has been used for finding similarity among motion trajectories. Class-wise global and local costs have been computed using DTW and their fusion has been done using Particle Swarm Optimization (PSO) to improve the classification rate. Experiments have been performed using two public trajectory datasets, namely T15 and LabOmni. The proposed method yields encouraging results and outperforms the state of the art techniques.