Image Tracking and Prediction of Offshore Tsunami with Particle Filters

In this paper, a particle filter based estimation and tracking method is developed as the auxiliary system to determine the trajectory of the float ball on offshore for tsunami detection. The computer vision techniques combined with the particle filtering algorithm have been successfully applied to the detection and recognition of a rapid moving object. For the nonlinear dynamics of tsunami, the trajectory tracking and detection of the tsunami is carried out by means of the camera image of floating balls on offshore. The round frame will be set up and the weight of each particle will be evaluated according to the assigned color values and the similarity degrees between particles are calculated following the prediction, updating, and resampling processes. Gradually the centroid of particles is recalculated after the resampling process that will close to the true object trajectory. Compare the updated trajectory to the reference level, an abnormal rise of the sea level can be detected and the tsunami warning can be issued such that people on coastal would be evacuated in time. Our simulation results show that the proposed algorithm is effective in detection and tracking the approaching waves of rapid rising.