Trajectory-Based Scene Understanding Using Dirichlet Process Mixture Model

Appropriate modeling of a surveillance scene is essential for the detection anomalies in road traffic. Learning usual paths can provide valuable insight into traffic conditions and thus help identifying unusual routes taken by commuters/vehicles. If are learned nonparametric way, manual interventions marking be avoided. In this paper, we propose an unsupervised method to learn frequently used from tracks moving objects Θ(kn) time, where k denotes number n represents tracks. proposed method, temporal dependencies considered make clustering meaningful using temporally incremental gravity model (TIGM). addition, distance-based learning makes it intuitive estimate parameters. Further, have extended TIGM hierarchically as dynamically evolving (DEM) represent notable dynamics scene. The experimental validation reveals that quickly without prior knowledge about ( k). We compared results with various state-of-the-art methods. also highlighted advantages over existing techniques popularly designing monitoring applications. It administrative decision making control at junctions or crowded places generate alarm signals, if necessary.