Contour-feature-based 3D texture reconstruction method for a constant-diameter straight pipe
Jiasui Yao, Xiaoqi Cheng, Haishu Tan, Xiaosong Li · Measurement Science and Technology · 2025
Abstract This paper proposes a contour-feature-based 3D texture reconstruction method for a constant-diameter straight pipe (CDSP), aiming to address the challenges of 3D texture reconstruction in tasks such as pipeline inspection by unmanned aerial vehicles, cable surface defect detection and digital twin modeling of tubular structures. Initially, the method calculates the pose of the pipe axis through the perspective projection model for a single view of a CDSP to achieve 3D geometric reconstruction of the pipe. Subsequently, it utilizes the spatial angular constraint formed by the camera optical center, spatially discrete points and the center of the cross-sectional circle to perform visibility filtering. Combined with the camera perspective projection imaging model, it completes the extraction of grayscale information from visible points on the pipe surface and the reconstruction of 3D texture. A simulation experiment is conducted using the CDSP simulated image with a checkerboard texture, and discrete sampling is performed with a central angle of 0.1°. The chessboard corners extracted from the projected image obtained by projecting the 3D texture are consistent with those extracted from the simulated image, and the average reconstruction time is 0.648 s. In physical experiments, an image of a CDSP with chessboard texture was captured. Discrete sampling was also conducted using a central angle of 0.1°. The average deviation between the chessboard corners extracted from the projected image and those from the captured images was 0.400 pixels, with an average reconstruction time of 0.452 s.