A fully automatic and rather flexible mobile vision 3D coordinate online measurement system is presented and analyzed. The system is composed of a high resolution CCD camera, retro-reflective feature points, and an external orientation device. The feature points, which fixed on the measured object and external orientation device, are imaged by camera in several locations and orientations. The initial external locations and orientations of the camera relative to the fixed external orientation device are determined by pose estimate algorithm, and the feature points coordinates, i.e. measured points coordinates on the object, are calculated by bundle adjustment algorithm based on collinear constraint. To improve the measurement accuracy, the locations of the imaging feature points are determined by bilinear centroid sub-pixel algorithm. The effectiveness of the proposed system has been tested by experiments.