Abstract: In recent years,crowd behavior analysis has become one of the hottest research points in computer vision.The main application is intelligent surveillance,human-computer interaction,intelligent household,video retrieval and other fields.It is a primary research goal to analyze behavior,understand crowd in video,execute motion detection,match and model the moving target when input video sequences.This paper summarizes research status,and the state-of-the-art of crowd behavior analysis.Firstly,the database for the current crowd behavior analysis is briefly introduced and compared.Then we classify the algorithm into feature-based and model-based ones according to the difference of the emphasis of crowd behavior analysis core algorithm,meanwhile we subdivide and compare the methods on the basis of respective characteristics of each category,and analyze their advantages,disadvantage and practical crowd scene of the representative algorithm in each category.Finally,we discuss the difficulty and challenge as well as the future research trends of crowd behavior analysis.