SIA OpenIR  > 工艺装备与智能机器人研究室
Alternative TitleDimensional Inspection of Automobile Sheet Metal Parts Based on Computer Vision
Thesis Advisor夏仁波
Keyword钣金件检测 点云拼接 圆孔检测
Degree Discipline模式识别与智能系统
Degree Name硕士
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Other AbstractSheet metal parts are widely used in the automotive industry because of their light weight and can be processed into complex parts. In order to ensure the processing quality and assembly quality of sheet metal parts, their dimensional parameters must be inspected. Generally, enterprises use specific fixture equipment to inspect mass sheet metal parts. But once the model of the sheet metal part is changed, the original fixture equipment cannot be reused, which causes waste and puts economic burden to the enterprises. The coordinate measuring machine can be adapted to the detection of different models of sheet metal parts, but due to its large size and slow inspection speed, the coordinate measuring machine cannot be used to inspect mass parts. Therefore, enterprises have an urgent need to achieve the rapid, high-accuracy and high-flexibility inspection of sheet metal parts. This paper has studied the dimensional inspection of sheet metal parts based on computer vision and finished the inspection by analyzing the point clouds and images of the sheet metal parts. The main contents of the paper are as follows: (1) To obtain a complete point cloud of a sheet metal part, it involves point cloud stitching. Error exists in the stitching of two point clouds, and when more or larger point clouds are stitched, the error will be accumulated. In order to reduce the error, this paper proposes a method which is call binocular sparse bundle adjustment. By minimizing the reprojection errors in multi-view binocular images, the stitching error is reduced 75%. (2) Circular hole is an important feature on sheet metal parts because it is often used as benchmark in processing and assembly. In a binocular system, it is challenging to reconstruct the hole because it is difficult to accurately find the edge of the hole in images as well as the exact correspondence of the hole edge point extracted from the left and right images separately. This paper proposes a method for high-accuracy hole inspection. In this method, the edge of the hole in an image is extracted by finding highly trusted edge points, and the epipolar geometric constraint is introduced to find the hole edge in the binocular images. The method allows to inspect holes with various diameters and thicknesses from different views, and the root-mean-square error is less than 0.05mm. (3) In order to achieve the inspection of sheet metal parts, this paper develops a binocular system and a software, and the standard inspection process is also given. Using a point cloud processing software developed by our laboratory, the inspection of sheet metal parts’ holes, faces and sides is achieved. The average error of inspection is less than 0.05mm.
Contribution Rank1
Document Type学位论文
Recommended Citation
GB/T 7714
苏润. 基于视觉的汽车钣金件几何参数检测[D]. 沈阳. 中国科学院沈阳自动化研究所,2020.
Files in This Item:
File Name/Size DocType Version Access License
基于视觉的汽车钣金件几何参数检测.pdf(27373KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[苏润]'s Articles
Baidu academic
Similar articles in Baidu academic
[苏润]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[苏润]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.