Combinatorial Optimisation and Linear Prediction Approaches to Rain Cell tracking

In this paper two complementary approaches for object motion tracking are presented. The objects considered, highly deformable, are rain cells issued from weather radar image sequences. The ørst approach consists in formulating the problem as an object-to-object assignment one, treated as a combinatorial optimisation problem and solved by ad hoc approximate heuristics. The second approach aims at implicitly identifying the motion components and predicting their future evolution by linear extrapolation. Experimental results are presented, pointing out the successful tracking of rain cells, possibly leading to the prediction of heavy rainfall. Keywords: motion analysis, hydrology, radar images, combinatorial optimisation, simulated annealing, tabu method, assignment. 1. INTRODUCTION Digital image sequence processing techniques play an increasing role in the analysis of time-dependent physical phenomena and their evolution. The problem considered in this paper is the study of rain cell..