Font Size: a A A

Research On Moving Objects Detection And Tracking Based On Adaptive Gaussian Mixture Model

Posted on:2018-12-03Degree:MasterType:Thesis
Country:ChinaCandidate:J M ZhengFull Text:PDF
GTID:2348330569986295Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
The detection and tracking of moving objects is a key technique in the research field of intelligent video surveillance,and thus has a promising future in applications.However,a series of external interference can seriously affect the results after the process of detection and tracking of moving object,such as illumination changes,the shadow of the object,the occlusion between the objects and so on.Therefore,the research of moving object detection and tracking is a hot topic in the field of intelligent vision.This thesis reviews the current research situation of the moving object detection and tracking algorithms,and briefly introduces the related basic theory and technology.On this basis,aiming at some prominent problems in current researches and applications,this thesis conducts a further study on the moving target detection and tracking algorithm.To solve the problems that the traditional Gaussian mixture model has poor adaptability to scene changed and poor ability to detect complete moving objects,and is sensitive to the illumination changes,an improved algorithm is proposed for moving object detection based on Gaussian mixture model.The algorithm first improves the traditional three-frame difference method,and use it to extract the roughly region of moving object,which uses the dynamic segmentation threshold and edge detection technology to effectively solve the problems of the illumination changes and the discontinuity of the object edge.Next,the number of mixed Gauss model in the updating process of Gaussian mixture model is chosen adaptively.In the end,the HSV color space is used to remove the shadow and the whole moving object is detected.The experimental results show that the proposed algorithm has better performance in real-time,accuracy and adaptability when compared to existing similar algorithms.As for the problems of low accuracy and poor stability of moving object detection and tracking algorithm,this thesis proposes an improved algorithm for moving object detection and tracking based on adaptive Gaussian background modeling and Kalman filter.First this method changes the learning mechanism of traditional weights in the background model update,so that the background model can be more effective.Next consider the motion information and the centroid position information of the moving object as the initial input data of the Kalman filter,and then the Kalman filter can begin to track the object.Finally,in the process of updating,the Kalman filter observes the noise covariance matrix with the state change of the moving object.When the moving object is blocked,the prediction function of the Kalman filter is used to keep track of the object.When the object is reappeared,the moving object is judged by the color feature,and the algorithm will keep tracking if the object is determined as the same one.The theoretical analysis and experimental results show that the proposed algorithm can detect and track the moving object,and has a better performance in both accuracy and stability.
Keywords/Search Tags:Object detection, Object tracking, Gaussian mixture model, Kalman filter, Shadow removal
PDF Full Text Request
Related items