To deal with the problem that traditional trajectory similarity algorithms always can’t distinguish noise and dissimilarity in trajectories, this paper proposed a new method to calculate the similarity between trajectories based on dynamic time warping(DTW) algorithms and normalized the final result for easy comprehension and can be used to rank multiple pairs of trajectories conveniently in data mining applications. The algorithm progress was also optimized. The experiment conducted on dataset sampled from real life indicated that this method is robust to noise and abnormal points, and does not have any assumptions on the sampling rate, it also works well even we only got partial trajectories. It also shows that our method has great advance in distinguishing the similar and dissimilar part of the trajectories, it also works when the trajectories are sparse sampled.