Abstract: The principle of constructing independent vector sets using Gram-Schmidt orthogonal transformation is proposed. The method of setting independent baseline and independent double-difference ambiguity data sets refers to the idea of vertex incidence matrix and vector weight. The three ways to build double-difference ambiguity data sets are discussed. The computing process of selecting independent baselines and independent double-difference ambiguities is provided. An example of orbit determination using global GPS track nets is tested. The result shows that the orthogonal transform algorithm is effective.