A three-dimensional (3D) data registration method based on planar circular target is proposed to solve the problem of 3D data registration in the vision measurement of large-scale object surface profile. With the planar circular target placed in front of the measured object, 3D vision system takes photographs of planar circular target at adjacent measurement positions. Then the planar circular target is used as an intermediary, and registration of 3D measurement data from adjacent measurement positions is achieved by establishing correspondence between edge points in circle. As shown in experiments, the data registration root mean square (RMS) error of x, y and z, with considerable accuracy of the planar grid target method, is 0.06, 0.03 and 0.134 mm respectively. When the measured object is partly occluded, the data registration RMS error of x, y and z is 0.06, 0.03 and 0.168 mm respectively. While the method of planar grid target fails to registration because of mismatch of 3D grid data.