Multi-Targets Tracking Association Based on Online Sequential ELM

With the clutter and dense routes of maneuvering targets such as airplanes, their trajectories are difficult to track and are easily disconnected. In this paper, we propose a method for identifying disconnected trajectories based on Multiple Model Nonlinear Smoothing Gaussian Probability Filtering algorithm (MM-SGM-PHD) and Online Sequential Extreme Learning Machine (OS-ELM), which is fast, accurate, and effectively avoids identifying false tracks as target tracks. First, the MM-SGM-PHD method is proposed to suppress clutter and track multiple targets. Then, based on the Kullback-Leibler divergence, a representative track feature is selected to estimate the disconnected track. Finally, the OS-ELM is employed to train and test the average speed of the track, the average acceleration on the x and y-axis, and the distance difference of each segment to determine whether the track segments are the same. Simulation experiments are given to verify the effectiveness of the proposed method. Moreover, the proposed method is more suitable for track association engineering applications than traditional tracking and association methods.