According to the low measuring accuracy from a flexible coordinate measurement machine, the error correction and parameter calibration methods for the coording measarement machine were researched. An improving genetic algorithm based on optimization least square method was proposed to implement the parameter calibration. The kinematic model and the error model of the flexible coordinate measurement machine were established by Denavit-Hartenberg(DH) method. Firstly, a variable factor was used in least square method to evaluate the convergence speed. Then, the suboptimal parameter was regarded as the initial population of optimized genetic algorithm while the variable factor became steady. Finally, the improving genetic algorithm was used to search and calculate to obtain the optimal parameter and the parameter calibration was finished. The experiment shows that the proposed algorithm has fast convergence speeds and good robustness.