Research and Simulation of Abnormal Image Detection Method for Crowd in Sensitive Area
Jiangfen Wang · Jisuanji fangzhen · 2015
The paper proposed a method of crowd abnormality detection in sensitive area based on the behavior analysis. According to the behavior type of crowd,the gradual evolusion of polynomial distribution along the time direction was designed. According to the type of behaviors,the mixed proportion of different behaviors was calculated.The some simple action types can be obtained according to the mixing ratio of different types of behavior,and all time- space words were obtained based on the related behavior types. The conditional probability was calculated by using the L* V matrix,to get the hyper- parameters,variables and the joint probability distribution of parameters. The Gauss radial primary function was taken as the kernel function,and using the to one classification algorithm to make the joint probability distribution quantizing into the histogram distribution,and input to the SVM classifier after normalization. According to the abnormal data training model,through the test data,the judgement of population abnormality in the sensitive area can be achieved. The simulation results show that the proposed method has higher detection accuracy.