Measuring motion trajectories of particle swarms in flight

We propose a novel dual optimization framework for measuring motion trajectories of large swarms of natural particles, in which a continuous objective model is defined in form of energy to approximate faithfully the behaviors of moving targets. Following the Lagrange dual decomposition strategy, the framework distributes the optimization problem into simple subproblems, each of which also approximate different behavior of targets respectively. With this "realistic" energy, the proposed scheme can approximate the underlying posterior of problems faithfully, while avoiding discretization errors. The new framework will take advantage of the complementary natures of subproblems, which will reduce ambiguity significantly while evading error propagation. Our experiments involve challenging datasets and demonstrate that our method can achieve results comparable to other state-of-the-art approaches.