Font Size: a A A

Three-dimensional Reconstruction Of Refractory Bricks Based On Binocular Stereo Vision

Posted on:2024-09-27Degree:MasterType:Thesis
Country:ChinaCandidate:W C WuFull Text:PDF
GTID:2531307142981119Subject:Electronic information
Abstract/Summary:PDF Full Text Request
Refractory bricks are building materials suitable for high-temperature and high-pressure environments.Due to the possibility of deformation and missing corners during the production and demoulding process,their use can be severely affected.Therefore,it is necessary to inspect refractory bricks.Currently,refractory brick inspection is mainly done manually by relying on experience,which is subject to significant subjective influence and cannot establish a unified inspection standard.Moreover,a large number of repetitive inspections can easily lead to misjudgment.A refractory brick 3D reconstruction system based on binocular stereo vision can obtain the 3D size information and surface defect information of refractory bricks and use this3 D information for automatic inspection,avoiding the shortcomings of manual inspection.Due to the insufficient texture information on the surface of refractory bricks,holes and other situations are easily formed in the 3D reconstruction process.Therefore,researching the 3D reconstruction of refractory bricks can improve the 3D reconstruction effect in a targeted manner and obtain more accurate 3D information.This article takes refractory bricks as the research object for 3D reconstruction,analyzes the basic process of binocular stereo vision,and focuses on researching stereo matching algorithms.The main work of this article is as follows:(1)The conversion relationships between four reference coordinates and two imaging models in the imaging process are introduced;the parallel binocular axis model and the principle of disparity are analyzed to explain the principle of obtaining three-dimensional spatial information from two-dimensional image information;the important concept of epipolar geometry in the binocular axis intersection model and stereo matching is expounded;the camera calibration and stereo rectification are discussed in detail.(2)Introduced the fundamentals of stereo matching,including algorithm classification,general process,constraint conditions,and similarity measurement functions;analyzed the difficulties and solutions of refractory brick stereo matching,providing ideas for subsequent matching algorithm research;elaborated on the commonly used algorithm detection platforms and algorithm detection standards,providing data references for subsequent algorithm research.(3)Introduced traditional semi-global stereo matching algorithms;focused on the ADCensus stereo matching algorithm.Based on the AD-Census algorithm,in response to the lack of surface texture in the stereo matching process of refractory bricks,the idea of vertically fusing image scales is introduced to improve the matching accuracy in weak texture and repetitive texture areas.At the same time,gradient transformation is added to the AD transform and Census transform for fusion,protecting the edge areas of the image and reducing matching errors in depth-discontinuous regions.(4)The principle of three-dimensional reconstruction of spatial points was introduced,a three-dimensional reconstruction system for refractory bricks based on binocular stereoscopic vision was designed,a hardware platform for the refractory brick three-dimensional reconstruction system was constructed,testing software for refractory bricks was developed,and the three-dimensional reconstruction of refractory bricks was realized.The test results show that the improved algorithm in this paper has an average PBM of6.69% in all regions,which is an improvement of 18.6% compared to the AD-Census algorithm.In the non-occluded regions,the average PBM is 5.07%,which is an improvement of 16.9%compared to the AD-Census algorithm.
Keywords/Search Tags:Binocular stereo vision, 3D reconstruction, Stereo matching, Refractory bricks, AD-Census algorithm
PDF Full Text Request
Related items