The problems of privacy preserving, trajectory data simplification and online processing attract considerable efforts from researchers in the area of trajectory data mining, unfortunately, it is difficult for traditional method to solve all these problems. This paper proposes an online similar trajectories mining method which can preserve the privacy of original trajectory data. The method first compressed and perturbed the original data based on random projection technique, then found the similar moving objects in each time segment by clustering the transformed data based on density, and finally it found similar trajectories by estimating that if the trajectories were similar for a long enough duration and estimated the similarity of trajectories using local sensitivity hashing. This avoided the intersection operation in traditional method and reduced the computation time. The experimental results show that this method can find similar trajectories effectively and reduce the cost of computation time. 