Moving object trajectory processing based on multi-laser sensing

There have been many researches on moving object detection and tracking. There are also great needs in trajectory analysis and scene modeling so that to provide higher knowledge to surveillance and ITS application for decision making. However in crowded environment, trajectory data sets obtained through online processing contain many broken, group and fragment ones, which degrades trajectory quality, affect the performance in trajectory analysis and scene modeling. A trajectory processing algorithm is developed in this research on a multi-laser sensing system that was developed in our previous work. It contains a trajectory association algorithm, where an interaction graph is built to represent the relationships of trajectories; a graph-based trajectory labeling algorithm; and an EM-based trajectory parameter optimization algorithm. Experiments are conducted using the data collected at an intersection in Beijing with promising results demonstrated.