Abstract: Aiming at the problem in improving the efficiency of crankshaft roundness error measurement, a method of visual measurement which was used in the detection of crankshaft roundness error based on convex hull was proposed. Firstly, the camera calibration and distortion correction of roundness visual evaluation system were carried out, and the designed acquisition system was used to obtain the rotating image of the crankshaft in real time. Then, the edge coordinates were extracted by noise removal and sub-pixel edge detection, and the discrete data points were obtained by three-dimensional reconstruction. Finally, a roundness evaluation model based on computational geometry technology was proposed, and the data in literature were used for simulation verification. The results of crankshaft roundness error of CMM are taken as standard values to compare and verify the proposed methods. The results show that the mean absolute error of machine vision inspection results compared with CMM is 3.39 μm. It proves that the visual evaluation method of roundness error has high accuracy, and can be applied to the evaluation of crankshaft roundness.