Font Size: a A A

Research On Fire Location Of The Large Space Building Based On Machine Vision

Posted on:2014-03-17Degree:MasterType:Thesis
Country:ChinaCandidate:K Z WangFull Text:PDF
GTID:2252330422455327Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of the national economy, the number of the large spacebuilding is increasing, because of their particularity and complexity, the conventionalfire detection and fire-extinguishing installation lose effectiveness for the fire of thelarge space building. In recent years, the image fire detection technology has been wellapplied in the large space building, the intelligent fire cannon technology has also beenapplied, on the basis of the fire localization of the traditional intelligent fire cannon, forthe shortcoming of its obvious hysteresis quality, the paper proposes the fire localizationbased on the parallel binocular stereo vision, which improves the real-time and accuracyof the fire localization, the application is very important.The technology of parallel binocular stereo vision and3-D informationreconstruction is used to determine the fire3-D coordinate of the large space building.Firstly, Zhang’s planar method is used to calibrate the internal parameters of cameras;Secondly, the SURF algorithm is used to extract flame image feature points, rapid graycorrelation coarse matching is used to match the feature points, and then estimating Fmatrix to restore epipolar geometry, the SURF algorithm is used to match the flameimage feature points in stereo matching. in order to improve the positioning accuracyand real-time, a new method is used to estimate the F matrix, it’s faster and moreaccurate, which is: first, LMedS algorithm is used to remove abnormal matching point;second, M-Estimators is used to dislodge positioning noise; the matching searchstrategy of SURF matching algorithm is improved based on its original advantage, withthe feature set groups, matching in turn according to the group, the matching accuracyand timeliness are improved. At last, the E matrix is computed by using the F matrixand internal parameters of cameras, the external parameters about the relative position of the parallel binocular camera are obtained by using the SVD of E matrix, then the3-D reconstruction technology is used to determine the three dimensional depthinformation of the fire.The experiments show that the proposed method of fire location in large spacebuilding not only ensures the requirement of real-time of fire localization, but alsoimproves the accuracy of fire localization, meeting the requirements of fire localizationof the large space buildings.
Keywords/Search Tags:Large Space Building, Parallel Binocular Machine Vision, F MatrixEstimation, SURF Matching Algorithm, Fire Localization
PDF Full Text Request
Related items