The accuracy of the reconstruction of 3D objects using vision-based techniques depends ultimately on an optimization technique known as bundle adjustment. However, direct application of bundle adjustment fails to make full use of the constraints inherent in the objects' motion. Most of the scanner systems available today use a turntable to rotate an object so as to capture images of the object from different view points. For these systems, the motion of the object is constrained to rotate around a fixed axis. By incorporating this constraint into the traditional bundle adjustment method, we show that a more accurate reconstructed model can be obtained. This "constrained" bundle adjustment method should be useful in the design of low-cost, yet accurate, 3D scanners for use in manufacturing and entertainment industry.