| In order to solve the problems of low efficiency and cost of human and material resources,this paper has analyzed the video data flow that can automatically detect target in the scene and analyze,identify and understand the target behavior based on detected target,and achieved intelligent monitoring and processing of the abnormal behavior in the monitoring scene.The specific work is as follows:Firstly,a framework that use several classical motion target detection algorithms has been constructed.With the consideration of performance and design time cost,the framework is used to select the best algorithm parameter combination to detect the moving target with the change of the scene.And run time can be adaptively adjusted about real-time input.This can effectively solve the problem of obvious difference in the detection effect of moving target in the different scenes,and enhance the versatility of the detection algorithm.Secondly,an algorithm based on fuzzy iterative self-organizing data analysis and clustering combined with histogram entropy is proposed to detect the abnormal behavior.The algorithm has used the method of fuzzy iterative self-organizing data analysis and clustering that obtains the video key frame,reducing the computational complexity.According to the classification results and the model of abnormal behaviors,three abnormal behaviors models,using the histogram entropy as a general feature to detect abnormal behavior has been built,and compared to Histogram of Oriented Gradient with Support Vector Machine for abnormal behavior detection algorithm.Thirdly,this paper has analyzed moving features of the three abnormal behaviors in different scenes,and obtained features parameters of the abnormal behaviors,setting the three abnormal behavior determination criteria,and the rules that the model ofdetecting abnormal behavior has stored in the database.Then an abnormal behavior rule base has been established.An abnormal behavior rule matching process is given.analyzed Whether the detected abnormal behavior matches the rules is analyzed.The experimental result shows that the rules of abnormal behavior can identify the specific abnormal behavior.In summary,the adaptive detection algorithm framework constructed in the intelligent video monitoring system has solved the problem at which the common detection algorithm is not strong.An algorithm of fuzzy iterative self-organizing data analysis and clustering combined with histogram entropy is proposed,which has higher detection accuracy than Histogram of Oriented Gradient with the Support Vector Machine.And the three abnormal behavior rule bases can match three kinds of abnormal behaviors.This algorithm can effectively improve efficiency of monitoring,and will have practical application value to maintain the stability of social security. |
