| Title |
Detection Method for Laser Grid Intersections on a Cylindrical Surface |
| Authors |
김민범(Min-Beom Kim) ; 박채호(Chae-Ho Park) ; 최강선(Kang-Sun Choi) |
| DOI |
https://doi.org/10.5573/ieie.2025.62.8.41 |
| Keywords |
Laser grid; Textureless cylinder pose; Cross point detection; 3D reconstruction |
| Abstract |
To obtain the data required for 3D pose estimation, we propose a method that projects a laser grid pattern onto a cylindrical surface and estimates the positions of the cross points. Through preprocessing steps such as Gaussian blurring and adaptive thresholding, horizontal and vertical laser line regions are extracted from the image, and initial cross points are obtained by calculating the centroids in the overlapping regions of the two lines. Subsequently, false detections caused by excessive brightness variations and noise are removed, and the valid cross points forming each laser line are fitted using a second-order polynomial model to refine the final cross points. Compared to conventional methods that rely on brightness information for correction?which suffer from reduced accuracy due to the large, periodic brightness changes occurring when laser lines intersect perpendicularly to form a grid?the proposed approach eliminates dependence on brightness variations by performing an initial binarization and utilizing only the positional information of the initial cross points to estimate the second-order polynomial model, thereby achieving more accurate cross point position estimation. Experimental results confirmed that the obtained cross point data significantly reduces the reprojection error compared to conventional Hessian-based methods. |