Abstract: Limitations of traditional roundness measurement methods for micro cylindrical components, including susceptibility to clamping errors, eccentricity and alignment deviation, were addressed by introducing a non-contact ultra precision coordinate scanning method for uncertainty evaluation. The method was developed based on the principle of segmental linear scanning and stitching, and an experimental platform composed of a chromatic confocal sensor and a linear displacement stage was constructed to measure a cylindrical workpiece with a diameter of 0.25 mm. The cross section was divided into eighteen segments for arc coordinate acquisition, and the roundness profile was reconstructed using least squares fitting and cross-correlation compensation. Uncertainty was evaluated through Monte Carlo simulation based on the reconstructed profiles. Experimental results show that under filtering scales of 50, 150 and 500 UPR, the average roundness values are 0.025 85 μm (standard deviation 0.003 43 μm), 0.100 31 μm (standard deviation 0.020 16 μm) and 0.236 89 μm (standard deviation 0.040 39 μm), respectively. The corresponding average radius values are 0.250 39 mm (standard deviation 0.000 95 mm), 0.250 42 mm (standard deviation 0.000 94 mm) and 0.250 44 mm (standard deviation 0.000 96 mm). The uncertainty evaluation indicates that under the 150 UPR filtering condition, the expanded roundness uncertainty is less than 0.0035 μm and the expanded radius uncertainty is approximately 0.018 μm. High precision, stability and reliability of the method are demonstrated, providing an effective solution for geometric accuracy evaluation of micro-scale cylindrical components.