Study On Multi-target Detection And Tracking Technology In Intelligent Monitoring System

With the rapid development of the city economy,our country is speeding up the pace of constructing "safe city" and has invested a lot of funds in the field of security monitoring,which has promoted the development of intelligent monitoring technology.Despite the emergence of new intelligent monitoring solutions,there are still many technical difficulties to be solved in this field.This paper mainly focuses on 3 key technical problems in intelligent Monitoring system:moving target detection,multi-targets real-time tracking and moving target intrusion behavior discrimination,and puts forward some improved methods.The specific research contents are as follows:Detect moving target.In this paper,the mixed Gaussian background modeling is used to obtain the binary initial extraction image of moving target in the surveillance area,but the initial extraction image is mixed with the interference of holes,small gap disconnected region,isolated light spots and cast shadows.The water-filled method is adopted to extract the connected region and remove the isolated spot noise.A shadow removal algorithm based on centroid estimation and shadow casting density is proposed to eliminate the interference of the moving target’s cast shadows.In this method,the region of interest(ROI)is determined from the binary moving object preliminary extraction diagram,and its centroid position is calculated to judge whether there is shadow interference in this region.If this region exist shadow interference,a shadow projection density histogram is established for this region,then maximum between-class variance method is used to determine the trough position of the histogram,and the shadow is removed by this boundary.Track multi-targets.A multi-target tracking algorithm based on inter-frame identification matrix is proposed by this paper.The algorithm uses the feature that the images between adjacent frames of the same moving target will certainly overlap to construct the inter-frame identification matrix.The motion state of each target is judged by the row and column scan matrix,then the number of the target and tracking strategy are determined by its motion state,and their motion trajectories are drawn.Eventually the real-time tracking of multiple targets in the video sequence is realized.Identify intrusion behavior.An intrusion behavior discrimination algorithm based on vector cross product is proposed by this paper,which realizes the intrusion behavior and direction discrimination of moving targets in the monitoring scene.Experimental results show that the moving target detection algorithm proposed by this paper can effectively remove the shadow interference and accurately extract moving targets.Multi-target tracking strategy based on inter-frame identification matrix can track multi-targets appearing in the video sequence in real-time and accurately.The intrusion discrimination method can effectively distinguish the intrusion behavior and the direction of the moving target.