Road segment partitioning towards anomalous trajectory detection for surveillance applications

Recently, the low cost and high availability of location acquisition technologies has significantly increased the demands for online anomalous trajectory detection. It is being used in social as well as commercial areas to provide human life care applications like healthcare, theft protection and taxi fraud detection. However, anomalous trajectory detection is still a challenging problem. The main complications involved in it, are inaccuracy in obtaining trajectory traces and evaluation of partial anomalous trajectories. In this study we contribute towards resolving these complications by proposing a novel method of Road segment Partitioning towards Anomalous Trajectory Detection (RPat). Our proposed method partitions the trajectory on the basis of road segments. Then, these sub-trajectories are evaluated, independently based on contemporary behavior of moving objects to accurately analyze the trajectories that possess abnormal behavior at any intermediate parts. The evaluation score of each sub-trajectory is aggregated to reflect the attitude of an overall itinerary as anomalous or regular. Further, the accuracy in the reconstruction of trajectories is achieved by plotting the itinerary traces on real world road maps. Experimental studies are conducted on real datasets and an accuracy of more than 81% is achieved.