The deformation in the aluminum hose tail during manufacturing will lead to a low success rate in the subsequent casing process. Aiming at the problem, it described an optimization model for searching the minimum area circle center with the goal of minimizing roundness error in this article. An improved roundness error evaluation method was also proposed that combined the geometric structure of minimum zone circle (MZC) with grouped particle swarm optimization (GPSO) algorithm. To evaluate the roundness of the aluminum hose, one approach based on machine vision was employed to obtain the sub-pixel edge location and subsequently roundness error was evaluated through the improved algorithm. The evaluation value were compared with those obtained through the least squares circle (LSC) and particle swarm optimization (PSO). The results show that there is the highest accuracy in the proposed algorithm where its computational efficiency is approximately 9 times that of PSO. To verify the effectiveness of the ameliorated algorithm, By comparing the evaluation results of the visual measurement in presented algorithm with those of the 3D scanning measurement in the relevant software, the error of the visual measurement relative to the 3D scanning measurement is less than 8%, which fulfills the requirements of the enterprise for the efficiency and accuracy of roundness detection for the aluminum hose tail, and verifies the effectiveness of the presented method.