Font Size: a A A

Research On Background Extraction Algorithm In Video-based Moving Object Detection

Posted on:2012-04-13Degree:MasterType:Thesis
Country:ChinaCandidate:R FuFull Text:PDF
GTID:2248330395955424Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the increment of the national highway mileage and the increasingly severeurban traffic congestion, the intelligent transportation system, as an important means ofsecurity and defense, has been paid more and more attention to by people. In generally,the traffic monitoring subsystem of intelligent transportation system consists of movingtarget detection, classification, tracking and video analysis, and several otherfundamental aspects of composition, which is the basis of moving target detection.In this paper, we objectively analyse several typical moving object detectionalgorithms. And after that, the background subtraction is determined as the algorithmwe study. And the key of background subtraction is the background extraction andbackground updating.For the computational complexity of the traditional background extractionalgorithm based on Gaussian mixture model, an improved algorithm is presented whichis based on the neighboring relativity. The algorithm establishes Gaussian mixturemodel through the interval effective pixels instead of the traditional point-by-pointpixels in background extraction. Experiments show that the proposed method reducesthe computational complexity of the video sequences processing and raises efficiency.For the background learning inefficiency of the traditional background extractionalgorithm based on Gaussian mixture model, an adaptive method is proposed to adjustlearing rate of Gaussian mixture model according to the ratio of the changes from framedifference. Experiments show that the proposed method improves the efficiency of thebackground study.In practical application aspects, the design and implementation of the movingvehicle detection system based on video surveillance system are introduced.
Keywords/Search Tags:Moving Object Detection, Background Extraction, Gaussian Mixture Model, Neighboring Relativity, Adaptive Learning Rate
PDF Full Text Request
Related items