Multiple moving objects tracking for automated visual surveillance

Moving objects tracking is of great significance for automated visual surveillance. Conventional tracking algorithms, such as Kalman filter or particle filter, have shown the effectiveness and robustness in many practical applications. However, the Bayesian filter is not designed for tacking multiple moving objects. The difficulty is the data association between the measurements and the tracks. Tracking can fail due to the confusion of similar measurements from adjacent moving objects. This paper proposes an approach for multiple moving objects tracking. We formulate the measurement assignment process as a problem of finding the matching with the maximum weight in a bipartite graph. Moving objects are detected by background subtraction. We test our approach using public datasets. The experimental results demonstrate that our approach is able to track multiple moving objects correctly.