In order to calculate the unstructured T-spline fitting surface with low distortion and meet the fitting error threshold and fewer control points for any complex topology fitting domain, we presented a step-by-step solution method. First, a polycube structure with the same topology as the fitting domain was generated as the parameter domain, and the corresponding relationship between the surface to be fitted and the parameter domain was optimized through multiple re-parameterization processes, thus obtaining a low distortion mapping suitable for the generation of low fitting error spline surfaces. At the same time, with the local subdivision property of the unstructured T-spline, the region, which did not meet the fitting error threshold, was adaptively subdivided, and the low distortion spline surface meeting the fitting error threshold was obtained. Next, a simplification strategy of fitting surface was presented to delete redundant control vertices. On the basis of meeting the fitting error threshold and low distortion, redundant control vertices were deleted, and a low distortion unstructured T-spline fitting surface was obtained with less control vertices and bounded error. The effectiveness of this method was verified on various complex models. Compared with the latest methods, this method could attain lower parametric distortion with fewer control vertices.