Abnormal crowd behavior recognition based on KOD energy model
Duan Jing-jin · Jisuanji yingyong yanjiu · 2013
In view of the weak descriptiveness problem on the crowd behavior which was caused by the fact that the shelter had a strong effect on traditional features in the abnormal crowd behavior recognition process,this paper proposed an abnormal crowd behavior recognition method based on the KOD energy characteristics. Neglecting the local characteristics of crowds in mutual occlusion,this paper defined and extracted crowd kinetic energy,orientation potential energy,distance potential energy on the basis of motion intensity,consistency of moving direction,relative location of the targets respectively. Then it constructed the KOD energy characteristics to describe the changes of crowd motion states. At last,the behavior type was distinguished from each other by constructing the hidden Markov models. Experimental results on the PETS and UMN datasets show that the method can detect and recognize different crowd abnormal behaviors with 92 percent accuracy,which is higher than traditional optical flow features.