Abstract: Identifying video stream quickly from the network flow is the foundation and prerequisite of network video surveillance and management. According to the analysis of characteristics of network video streams, a fast recognition and classification algorithm of network video stream based on multiple pattern matching is proposed, which constructs the matching machine with the features of different video streams. The proposed algorithm could determine whether the network packet contains video stream and type only by one incomplete scan. The experimental results show that the proposed algorithm has higher efficiency than traditional protocol analysis method under the requirement of accuracy.