Laser & Optoelectronics Progress, Volume. 62, Issue 12, 1215007(2025)
Flatness Inspection Algorithm for Rail Vehicle Sidewall Panels Based on Point Cloud
Figures & Tables(18)
Fig. 1. Front-end data acquisition device. (a) Simulation diagram; (b) on site diagram
Fig. 7. Comparison of single frame point cloud data. (a) Standard single frame point cloud data; (b) single frame point cloud data with defects
Fig. 9. Sidewall panels 3D model drawings. (a) Standard 3D model drawing; (b) linear area with defects 3D modeling drawing; (c) curved area with defects 3D modeling drawing
Fig. 12. Filtering experimental results. (a) Point cloud of standard defective contours; (b) noisy contour point cloud; (c) bilateral filtering effect; (d) guided filtering effect; (e) improved guided filtering effect
Table 1. Summary of advantages and disadvantages of point cloud defect detection methods
View tableTable 1. Summary of advantages and disadvantages of point cloud defect detection methods
Method Advantage Disadvantage Based on point cloud outline Strong adaptability and fast processing speed Large limitations， different product selection thresholds vary Template based matching Suitable for standard parts， high accuracy Lack of standard models requiring reverse reconstruction and high complexity Based on local geometric features Strong generalization ability Poor universality， requiring a large number of samples Based on multimodal point cloud data High precision and strong generalization ability High processing complexity and the need for 2D-3D mapping relationships Based on deep learning Capable of handling irregular point clouds Difficulty in obtaining samples and lack of universality
Table 2. Comparison of evaluation indexes of filtering effect
View tableTable 2. Comparison of evaluation indexes of filtering effect
Filtering algorithm Noise standard deviation Entropy RMSE SNR Bilateral filtering 0.10 7.332 0.134 55.790 0.15 8.130 0.140 55.290 Guided filtering 0.10 7.368 0.119 56.868 0.15 8.159 0.122 56.468 Improved guided filtering 0.10 7.344 0.078 60.491 0.15 8.197 0.094 59.233
Table 3. Parameter comparison of linear area with defects
View tableTable 3. Parameter comparison of linear area with defects
Item Length Width Depth Length error Width error Depth error Defect 1 measured value 299.79 249.88 10.45 0.21 0.12 0.08 Defect 1 true value 300.00 250.00 10.54 ‒ ‒ ‒ Defect 2 measured value 299.03 299.07 0.81 0.97 0.93 0.01 Defect 2 true value 300.00 300.00 0.80 ‒ ‒ ‒ Defect 3 measured value 149.26 149.32 13.50 0.74 0.68 0.02 Defect 3 true value 150.00 150.00 13.52 ‒ ‒ ‒ Defect 4 measured value 399.09 149.06 7.91 0.91 0.94 0.09 Defect 4 true value 400.00 150.00 8.00 ‒ ‒ ‒
Table 4. Parameter comparison of curved area with defects
View tableTable 4. Parameter comparison of curved area with defects
Item Length Width Depth Length error Width error Depth error Defect 1 measured value 299.01 127.60 5.76 0.99 0.80 0.22 Defect 1 true value 300.00 128.50 5.98 ‒ ‒ ‒ Defect 2 measured value 698.99 159.03 7.29 1.01 0.97 0.22 Defect 2 true value 700.00 160.00 7.07 ‒ ‒ ‒
Tools
Get Citation
Copy Citation Text
Qi Sun, Lintao Huo, Haifei Xia, Yutu Yang, Bin Wu, Sheng Xu, Ying Liu. Flatness Inspection Algorithm for Rail Vehicle Sidewall Panels Based on Point Cloud[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1215007
Paper Information
Category: Machine Vision
Received: Nov. 15, 2024
Accepted: Jan. 2, 2025
Published Online: Jun. 23, 2025
The Author Email: Ying Liu (lying_new@163.com)
CSTR:32186.14.LOP242265
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20