Research On Human Flow Density Detection Algorithms Based On Contour Analysis

Currently,the detection of human flow density is the frontier research direction in the field of computer vision.In daily life,crowd gathering often brings abnormal events and unexpected incidents.The method of human monitoring and statistics of human flow density has low accuracy and high cost,so it can not detect potential safety hazards in time.Therefore,real-time intelligent video surveillance system can be installed in public places to dynamically observe the changes of human flow,and accurately warn passenger flow to avoid accidents.In this paper,aiming at the problem of human flow density detection,the real-time human flow density detection function with low false alarm rate and high accuracy required by the system is realized by combining the practical application of the intelligent detection system of video surveillance on the Thousand-mile Eye Platform.This paper presents an algorithm for detecting human flow density based on contour analysis,which overcomes the complexity caused by dynamic complex scenes and dense human flow.This paper focuses on contour analysis and carries out algorithm design.The main research contents include the following aspects:In order to improve the reliability of contour detection,this paper designs an edge detection method based on Mixture Gauss Background Modeling.Traditional foreground edge detection methods usually first detect the foreground,and then detect the foreground edge on this basis.In contrast,this paper directly establishes the edge background model,first extracts the edge of the video image,then combines with the method of Mixture Gauss Background Modeling to detect the foreground of the video image quickly and effectively,which overcomes the instability of the edge detection caused by the color or gray background modeling.The experimental results show that this method improves the robustness of the system,adapts to the changes of illumination conditions and noise,and improves the stability and continuity of foreground edges.In order to overcome the problems of edge detection error and false alarm caused by abnormal image quality,an image quality detection function based on feature point tracking is proposed in this paper.Firstly,feature points are extracted and tracked,then motion features of feature points are extracted,and SVM classifier is constructed to recognize and classify image quality anomalies,so as to realize the function of image quality detection.Theexperimental results show that the algorithm can be well applied to the detection of camera anomalies in various complex scenes,effectively reduce the false alarm rate,and has a high recall rate.Aiming at the problem of human density detection in dynamic complex scenes and dense human flow,a human density detection algorithm based on human contour recognition and tracking is proposed in this paper.Firstly,the contour features on both sides of human body based on foreground edge are extracted from video images,and then the contour motion features are extracted by contour tracking.On this basis,the reliable detection and recognition of human contour are realized.Finally,the pedestrian density in human flow is reflected by contour density,and the detection of human flow density is realized by multiple linear regression.The experimental results show that the proposed detection method based on contour analysis can achieve real-time detection of low false alarm rate and high accuracy under dynamic complex scenes and dense traffic conditions.The method proposed in this paper has been put into practical application in the Intelligent Detection System of Video Surveillance on Thousand-mile Eye Platform.