Remote Pipe Diameter Measurement from a single Image using Laser Scale Projection with a Depth Compensation Model

Abstract. The dimensional measurement of cylindrical components, such as pipes, is a critical demand for quality control in industrial inspection. Monocular (single-view) computer vision-based approaches offer a low-cost and highly flexible solution but face the fundamental challenge of scale ambiguity. This paper proposes a method for measuring pipe diameter that solves this ambiguity by utilizing the projection of two laser points onto the object from the beam-splitting of a single laser source. The system projects two points with a known and previously calibrated distance onto the object's surface, providing the necessary scale information (mm/pixel) for the image. However, the projection onto the curved surface of the pipe induces a systematic error, as the laser dots land at a different depth (Z) than the depth of the apparent edges of the diameter in the 2D image. The main contribution of this work is the development and validation of a curvature compensation algorithm that mathematically corrects this depth distortion. Experimental results demonstrate that the proposed method, by compensating for the curvature, enables precise diameter measurement with a simple monocular system. For a pipe with a nominal diameter of 165 mm, measured at 6 m distance, the bias in the estimated diameter was reduced from 1.5% to less than 0.2%.