Binocular Vision Reconstruction Algorithm And Application In Transformer Reconstruction

With the rapid expansion of power grid scale in our country and the automationlevel unceasing enhancement, construction of UHV is developed vigorously, which isbecoming a hot topic of power grid. In a long distance transmission line, transformersplay more and more important roles. In order to improve the security and reliability ofthe electric power production and transportation, transformer fault diagnosis isrequired. Nowadays, with the application of UAV patrol and the rapid development ofhandheld fault detection device, more and more new technologies are implemented.Under the background, remote digital video monitoring and image recognition systemarise at the historic moment. It can receive digital video signal collected at the remotescene, and transmitted to the monitoring center in real time. In the monitoring centerremote site monitoring, Control remote camera’s motion, and according to the need toanalyze the collected digital video image, processing and recognition.In order to reflect the state of the transformer more accurate and comprehensiveand improve the function of image recognition part of existing remote vision system,meanwhile, improve the system automatically matching speed and accuracy of targetimage under the complex weather conditions. In this paper, based on Scale InvariantFeature Transform (SIFT) algorithm and combined with nuclear line matching, theSIFT algorithm is improved. First of all,using the knowledge of the epipolar image,the two images of the same name nuclear line are found, matching on the epipolar ray,matching time is reduced. Secondly, after the match, get space three-dimensionalpoints from matching point. But application of the improved matching algorithm cannot reconstruct the original image, a small scale regional growth of parallax image isnecessary, then, transformer reconstruction images are obtained. in this paper, Ireduce the matching time and speed up the matching speed in the improved SIFTalgorithm, region growing algorithm is introduced in the parallax diagram, get the transformer three-dimensional images. The paper take transformer as experimentalobject, the experimental results show that the algorithm can reduce the time of imagematching, meanwhile, get the3d image of the transformer, has a good applicationprospect in the visual monitoring transformer.In this paper has certain theoretical significance and practical application valueto improve the reliability of remote video monitoring, image matching andrecognition system, although expand its application space. The development of thesubject will help to further improve the information level of the power system and thereliability of the UAV patrol automation level, make important contributions to thedevelopment of electric power system in our country.