Abstract: A real-time object track algorithm based on particle filter and adaptive template update is proposed.On the framework of particle filter,a mixture Gaussian model is introduced to be an effective target description,which is composed of three components and online updated to solve the template update problem by using an incremental EM algorithm.Experiment results show that the algorithm can track infrared targets under variational illumination,especially for fast varying appearance and pose and local occlusion,furthermore,it is simple enough to be a real-time object tracking algorithm without particle resample.