To deal with the heavy-tailed issue of the video object with abrupt turns，a particle filter tracking algorithm based on switching observation models is proposed. According to the presented switching rule，to improve the tracking precision，the observation likelihood of Gaussian distribution is used when the object moves steadily，and the observation likelihood of multivariate Laplace distribution is adopted to surmount the heavy-tailed issue when the object turns abruptly. The simulation result shows the algorithm is robust and effective at abrupt turning object tracking.