Detection Of Salient Crowd Motion Based On Repulsive Force Network And Direction Entropy

Intelligent monitoring is an important application direction in the field of artificial intelligence,and crowd behavior analysis is a research hotspot in the field of intelligent monitoring.It mainly analyzes the movement state of the crowd by extracting the characteristics of pedestrian behavior in the video image,so that the system can detect the operation state of the crowd worthy of attention,so as to alert or provide reference for the staff,so as to reduce accidents and dangerous events Occurrence of pieces.In crowd movement,significant crowd movement usually represents the behavior inconsistent with the mainstream pedestrian movement.For video monitoring,these movement behaviors deserve more attention.Detection of significant crowd movement also has great application value in crowd video analysis.In the crowd,different sports pedestrians will have repulsive force in order to avoid collisions.For the significant sports pedestrians who are inconsistent with the mainstream crowd movement,their repulsive force with the mainstream crowd will be great.At the same time,the movement direction of different pedestrians may be different,while the movement direction of significant sports people is more chaotic,so the corresponding direction entropy is larger.Combined with the characteristics of repulsive force and direction entropy,this paper proposes a population-based motion detection algorithm based on repulsive force network and direction entropy.The main research contents and innovations are as follows:1.The crowd weighted network is constructed by calculating the repulsive force between velocity vectors.The pyramid optical flow algorithm is used to calculate the velocity vector field of crowd movement,and the relationship between the velocity vectors in the field is analyzed with the thinking of network.Each velocity vector is regarded as a node in the network.The repulsion force between nodes is calculated by the repulsion force formula,and the module value of repulsion force is taken as the connection weight between nodes,so as to build the crowd weighting based on repulsion force Complex network model.The repulsion network is represented by correlation matrix.The static geometric properties of repulsion network are analyzed,and the strength characteristics of nodes are extracted,and the strength matrix(field)of nodes of repulsion network is obtained.2.Optimize the node strength field by the motion direction entropy.The velocity vector motion direction is graded.According to the definition of Shannon information entropy,the motion direction entropy is calculated for the velocity vector,and the motion direction entropy matrix describing the crowd is obtained.The feature fusion of the node strength field and the motion direction entropy is carried out.The node intensity field is optimized by analyzing and comparing the normalized addition method,and then the large-scale crowd behavior is analyzed by the numerical value change in the optimized node intensity field to detect large-scale crowds.Significant movement of the scene.The two scenarios of population retrograde and unstable crowd movement were tested significantly.The simulation results show that the optimized node intensity field can detect the significant movement of the crowd,and the other methods are used to detect the instability of the crowd in different scenarios.The results were compared and compared with the two characteristic parameters of accuracy and recall.The results show that the detection results of the proposed method are closer to the true value.