| The research content of this paper involves digital image processing,pattern recognition,computer vision and 3D modeling.Quickly and accurately detecting the monitoring target is the premise and foundation of the research.Accurately analyzing the basic behavior of the monitoring object is the key to the article.The identification of the behavior and the three-dimensional virtual monitoring of the monitoring screen are the application significance of this paper.This article uses the laboratory surveillance video of our school as a data source.The research content is divided into the following five aspects:(1)Theoretical analysis and experimental comparison of three commonly used moving target detection algorithms,namely cumulative difference method,optical flow method and background difference method.According to the advantages and disadvantages and complexity of each algorithm,combined with the characteristics of indoor character surveillance video,the background difference is selected as the main research method.(2)Discuss the mainstream background modeling method,and use the background modeling method based on the mixture Gaussian model for the characteristics of the indoor periodic changes of the background of the indoor surveillance video.The principle,calculation process,and experimental verification and result analysis of the algorithm are described in detail.Combined with the mathematical morphology processing and the shadow suppression method in the HSV color space to improve experiments,the interference of noise and shadow on the foreground detection results is effectively reduced.(3)Identify the target human body with the 7-joint point skeleton parameter model,which reduces the complexity of modeling,facilitates real-time processing,and well maintains the basic characteristics of human motion.A fuzzy discriminant method was proposed to analyze the basic behavior of human body and four kinds of human behavior quantificational analysis were achieved.(4)Mapping the monitoring space in a plane rectangular coordinate system and fitting the experimental algorithm to calculate the position coordinates of the monitoring object in the two-dimensional plane.The monitoring object's dwell time is calculated based on the front and back coordinates of the monitoring object,the direction of movement of the monitoring object is calculated using the 4-domain method,the object's travel path is calibrated,and the behavior information of the monitoring object is accurately presented in the two-dimensional plane.(5)A three-dimensional virtual platform for indoor surveillance screens was constructed to use the basic description parameters of indoor space,the results of the monitoring object behavior analysis,and 3D Max and other software to construct indoor space and object and character model to generate a three-dimensional virtual model of the monitor screen in real time.Experiments show that the method proposed and applied in this paper can effectively detect indoor moving objects and behavioral attitudes,object positioning,travel trajectory,and attitude staying time in real time.The monitoring scenes and objects are visually presented in two-dimensional and three-dimensional visualizations to protect the privacy of the monitoring objects.The results of this study are used to monitor and detect objects in office and indoor workplaces,unattended indoor venues,home elderly,and children.Behavior analysis has certain reference value and application significance. |
PDF Full Text Request