The artificial circular targets are used commonly in optical measurements including camera calibration and 3D object reconstruction. In the present study, the researchers usually depend on experience and intuition to manually determine the size of the circular targets, which is found to be important to the accuracy and efficiency of image position measurement. A larger target size may lead to a larger eccentricity error of the center of a circular target, while a smaller target is disadvantageous for image processing. Therefore, a suitable target size is required to guarantee a reliable image point measurement. In this paper, the problem of circular target size selection is investigated. Firstly, a theoretical model about the perspective imaging of a circular target is obtained. This mathematical model further leads to a novel and practical method for selecting a proper size of circular targets. Finally, experiments using circular coded targets are performed to validate the correctness and effectiveness of the proposed method. The experimental results indicate that the measured semi-major axis of the image ellipse corresponding to the circular target agrees well with its theoretical design value, and the existing differences are mostly within the tolerable range of the measurement error, as the maximum difference is estimated to be about 0.5pixel. Without any calibration parameters and high-cost equipment, this method shows significant potentials in various optical measurements.