Range data registration is a key issue of 3D optical measurement. The method based on markers has the advantage of cheapness, convenience and fastness. But the problem of error accumulation is still unsolved for multiple viewing fields. A fine registration method to register the multiple range data with non-coding markers is proposed. Image process and pattern recognition techniques are employed to detect the sub-pixel location of markers. The markers’ local coordinates in 3D space are computed from optical measurement system. Through the principle of rigid invariance of distance in 3D Euclidean space, the fine registration of two range data is completed. Moreover, a global iterative algorithm is proposed for accurate registration of multiple range data. Using this algorithm, the maximum error reduces from 0.123 mm to 0.076 mm. The experimental results illustrate the validity of this method. It can automatically fast register multiple range data and decrease error accumulation.