Research On Human Behavior Recognition Based On Image Processing

Intelligent monitoring has long been a hot research direction in the field of computer vision.It is the use of computer technology to analyze the video sequence so that the computer system can understand and perceive the world like the human brain.As an intelligent monitoring method,intelligent security systems are increasingly accepted and favored by people.In recent years,with the rapid progress of computer technology,people have made great progress in the field of intelligent monitoring,and obtained some very good research results.However,the current scientific research results still can not fully meet the application requirements of real life.In some key technologies,researchers still need to work hard to solve and study.At the same time,starting from the theory and reality,this paper makes an in-depth study on the key technologies of human behavior recognition based on video sequences.The research work of this paper mainly starts from the following aspects:(1)Through the comparative study of the commonly used image preprocessing methods,this paper chooses an adaptive median filtering method,and compares several popular filtering algorithms with the current popular ones to verify the adaptive method.The value filtering method can remove noise more effectively and preserve the details of the image well.(2)Researched some target detection algorithms that are commonly used in the academic field.After a large number of simulation experiments,after summarizing the advantages and disadvantages of various target detection algorithms,a hybrid algorithm that combines a hybrid Gaussian model and a continuous four-frame difference algorithm is proposed.It is verified through simulation experiments that good results have been achieved.expected result.(3)The study of clustering methods was carried out.After analyzing the actual clustering effect of K-means and its advantages and disadvantages,the existing K-means algorithm was improved,and the new algorithm was used to perform the attitude features.Clustering,the results obtained by the clustering process as HSMM parameters for classification.(4)Introduce the relevant theories and algorithms of the Markov model,based on the traditional Hidden Markov Models(HMM),and derive the hidden semi-Markov Models used in this paper,and apply HSMM to the field of behavior identification.It compensates for some deficiencies of the traditional hidden Markov model,such as the disadvantage of low recognition rate due to occlusion or other reasons.The new algorithm can more fully model human behavior.Through experimental verification,the HSMM model has a very good value in practical life.