Font Size: a A A

Feature Detection Technology Of Movement Target Based On Binocular Panoramic Vision

Posted on:2013-04-17Degree:MasterType:Thesis
Country:ChinaCandidate:C TangFull Text:PDF
GTID:2248330377958399Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Panoramic vision has become an Application method in Video monitoring and imagedetection with its advantage of full view in recent years. Along with the development ofresearch, capacity and speed for processing the internal information of panoramic visionimage is also strong.The analyses for internal image information of target environmentbecome more important. Binocular panoramic vision can also obtain the depth of fieldinformation in the basic of panoramic vision.First, analysis the structure of binocular vision system. Set up two kinds of binocularvision system which is suitable for panoramic view through the research for the camerageometry model and Fold reflection principle. Get a method of binocular vision system whichcan realize the design of panoramic view, this method can rapidly calculate the structuralparameters of the system.Secondly, design a kind of calibration methodfor panoramic vision system with thenonlinear characteristics. United the camera and the mirror into a system. Calibrate the systemto get the inside and the outside matrix parameters. This method avoided the errors causedcalibration after integration.At the same time using iterative fitting and the method of ballprojection to realize calibration for the panoramic view system quickly. The error is less thana pixel information units after the calibration through the experiment.Again, design a method which can obtain moving target fastly. This method combinesthe advantages of frame difference method and the background finite difference method, so itcan extracte moving objects in the sequence of images in real-time. Traditional SIFTalgorithm have the faults of testing time long and matching the efficiency lowly,so we set upthe characteristic information sampling accuracy and sampling number according to differentcharacteristics of the target environment to realize the extraction for SIFT effect fastly.Finally according to the imaging characteristics of binocular panoramic vision system,establish a binocular panoramic vision3d spatial calculation model.Then,we use the objectappear at specific angle,so we can take the three dimensional space problem into a2d spaceprogram to simplifie the3d calculating algorithm. According to the results obtained bymatching the characteristics, we can realize motion trajectory of the reconstruction andanalyse the characteristics of the movement.
Keywords/Search Tags:Binocular panoramic vision, Panoramic calibration, Feature matching, 3D reconstruction
PDF Full Text Request
Related items