Font Size: a A A

Research And Improvement Of Region Matching Algorithm In Binocular Vision

Posted on:2019-07-06Degree:MasterType:Thesis
Country:ChinaCandidate:M L ZhangFull Text:PDF
GTID:2428330542987797Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
In recent years,with the development of deep learning and the rise of artificial intelligence,computer vision has also played an irreplaceable role.Binocular stereo vision is an important branch of computer vision,which uses two cameras from different angles and gets the two 2D images.Disparity information can be obtained by stereo matching through the two images,and then the depth information of 3D scene can be got,which is widely used in position estimation,driverless technology,cultural relic recovery,fruit picking and so on.It has been widely concerned because of its simple and low cost advantages.Stereo matching is the core step of binocular vision,and the effect of matching will directly affect the application results.Therefore,the research of matching algorithm has always been a hot topic.Matching algorithms are mainly studied in this paper,the main research contents are as follows:1.The basic theories of binocular stereo vision and stereo matching are researched.The commonly used cost calculation algorithms and cost aggregation algorithms are researched and implemented,and also their advantages and disadvantages are compared.2.Aiming at the problem of image matching which is difficult to match in the stereo matching process,a stereo matching algorithm of low texture detection and cross-scale cost aggregation is proposed.In this algorithm,the low texture region of the images is detected firstly,and then the matching cost is calculated.Based on the results of low texture detection,the different scales are used to aggregate the matching cost for texture regions and low texture regions,respectively.By this way,not only the matching accuracy of the texture region is ensured,but also the error rate of the low texture region is reduced.The improved algorithm is then evaluated using the image pairs provided by the Middlebury site in order to verify the effectiveness of the improved algorithm.3.The 3D reconstruction is accomplished through the disparity map.The basic theories of camera model,3D reconstruction and texture mapping are studied and the triangulation method is used to complete the 3D reconstruction.In order to enhance the authenticity and recognizable of the objects,texture mapping is applied to the results of reconstruction.
Keywords/Search Tags:stereo matching, low texture matching, cross-scale aggregation, texture mapping, 3D reconstruction
PDF Full Text Request
Related items