Research On Moving Objects Detection Algorithm Based On GMM For Shadow And Background Adaptation

In computer vision and image processing,moving target detection has been an important research direction,moving target detection is the target acquisition and analysis of target tracking and target behavior based motion detection work,the use of a wide range,such as in computer vision system,intelligent monitoring technology and video compression and transmission.However,for different application scenarios,the lack of moving object detection method is the high complexity and poor stability of the algorithm,resulting in a lot of problems in the application and research of moving object detection.In this thesis,the method of moving object detection and background modeling is described in detail,and an improved method is proposed for the effect of the fixed rate of the shadow and mixed Gauss background model.For the above problems,the research work of this thesis:(1)the common algorithms of moving object detection are discussed.As for background modeling plays an important role in moving target detection,so this thesis focuses on the discussion of the Gauss background modeling,including a single Gauss background model and the mixed Gauss background model principle and performance comparison.(2)due to the large influence of moving target shadows of moving target detection,this thesis analyzes the causes and characteristics of shadow,and analyzes the scholars of the shadow detection algorithm of work before;and then analyzed to detect and remove the shadow of the methods and the advantages and disadvantages of different color space based on color HSV based on the method of shadow detection and elimination is proposed.(3)the principle of mixed Gauss model is discussed,and the basic algorithm is given.Analysis of the different parameters in the setting of Gauss mixture model modeling mixed Gauss model,the mixed background mixed Gauss model in the traditional model updating rate shortcomings of fixed,method of moving object detection by improving the update rate of implementation of the adaptive background change.The experimental results show that the proposed method is superior to the traditional methods in the accuracy of detection,and it can be used as a reference for the related research and application.