To achieve automated, high-precision in-situ diameter measurement of large-scale shaft parts, a robot-based measurement system integrating visual guidance and laser tracking is designed and constructed. The proposed system addresses the challenges of laser point guidance, alignment, and stable measurement under complex operating conditions, including significant initial pose deviations between the robotic arm and the workpiece, as well as highly reflective surfaces. By fusing semantic segmentation with temporal prior information, a robust target measurement region with a safety margin is established. On this basis, an image-based visual servoing (IBVS) pre-alignment strategy is designed to realize closed-loop guidance and real-time adjustment of the laser projection point, ensuring its stably and accurately entry into the target measurement region. Furthermore, a dual-edge joint weighted robust fitting and geometric solution method is employed to obtain the axial direction of the part and the surface normal at the contact point, and the end-effector posture is adjusted through iterative closed-loop control to achieve stable contact and reliable conformance between the roller and the workpiece surface. By utilizing a laser tracker to collect the 3D coordinates of a spherically mounted retroreflector attached to the roller as measurement points, the robotic arm is driven to perform multi-point data sampling within the measurement region. Combined with a cylinder fitting algorithm, this approach enables the automatic selection of measurement regions and diameter estimation across different or identical shaft segments. Experimental results indicate that the minimum absolute deviation between the fitted diameter fitted by the proposed method and the nominal diameter of the part is 0.108 7 mm, with the maximum absolute deviation not exceeding 0.240 0 mm. The standard deviation of the diameter across five repeated measurements remains below 0.180 0 mm, and the results show good consistency with the measured values obtained using a micrometer. The proposed method is suitable for automated, high-precision diameter measurement of large-scale shaft parts in engineering applications, demonstrating good engineering applicability and practical value.