Moving Object Tracking Based on Codebook and Particle Filter

In intelligent video surveillance, the key issues are the real time and robustness in multi-object tracking. We proposed a mixed tracking algorithm based on codebook and particle filter in this article. The algorithm used the codebook background model to detect the moving objects in complex scenes; Based on particle filter's special advantage in tracking non-linear, non-Gaussian moving objects, we used it for predicting and tracking objects, effectively resolving the interference of background color and the problem of object crossing and partly sheltered. The simulations show that this algorithm can steadily track the multiple moving objects in the outdoor scene, also can detect the abnormal behavior of the tracked objects.