The Algorithm Of Micro Motion Object Detection Based On ViBe And Multi-feature Extraction

With the rapid development of computer technology,a series of application fields such as intelligent video surveillance and analysis,remote video conferencing and teaching,and operator customer service video service have been developed.In these fields,the moving object detection and extraction of video image sequences has become a difficult research problem.Many methods have been proposed and improved by domestic and foreign scholars,but the extraction algorithm for micro motion object is relatively less,so the study of the micro motion object extraction algorithm in video image sequences has important theoretical and practical value.The micro motion object extraction algorithm in video image sequences was deeply studied in this thesis,and the research focus was mainly to improve the accuracy of micro motion object extraction.Firstly,this thesis proposed an improved ViBe algorithm with multi-feature fusion background modeling based on the spatial and temporal domain information.Secondly,another micro motion object extraction algorithm was proposed based on double ViBe background modeling and motion template.Finally,a micro motion object extraction and background replacement system was designed and implemented by integrating the two micro motion object extraction algorithms..The main work of this thesis as follows:1.This thesis studied the micro motion object extraction algorithm proposed by domestic and foreign scholars,and deeply studied the background modeling algorithm.Then some of the micro motion object extraction algorithms based on background modeling were experimentally analyzed.2.In this thesis,the analysis of the advantages and disadvantages of the micro motion object extraction algorithm based on ViBe background modeling was deeply studied and the ViBe algorithm was improved.The thesis proposed a micro motion object extraction algorithm of multi features background modeling based on spatiotemporal domain information.The foreground probability map of the transformed chromaticity information in YCbCr color was obtained by using the improved spatiotemporal domain texture operator.HSV shadow detection and foreground probability map were included in the target detection.The thesis improved the accuracy of the micro motion object extraction by improving the background matching and updating method,and the accurate replacement of the background was achieved.3.This thesis added time filtering for ViBe background samples,changed sample life cycle,and extended the number of samples to improve ViBe algorithm base on the analysis of the relevance of the video sequence images.And a micro motion object extraction algorithm based on double ViBe background modeling and motion template was proposed.The motion template was updated in real time by using the micro-target of the upper and lower background model.Then the object was detected by the object location and detection,and the algorithm was verified by experiment and data analysis.4.By combining the two kinds of micro motion object extraction algorithms proposed in this thesis,the micro motion object extraction and background replacement system was designed and experimented,and the functional modules of the system were tested by experimental video.At the end of this thesis,the research work was summarized and the future research direction and ideas were pointed out.