To achieve high-precision measurement of the complete three-dimensional (3D) shape of double-specular-surface objects, a global optimization method for system calibration is proposed. Firstly, a double-specular-surface shape measurement system is established using two direct phase measuring deflectometry subsystems without overlapping Field-of-View (FoV), which fully covers the measured FoV of the double-specular-surface object. Secondly, traditional methods are applied for depth calibration, lateral calibration of each subsystem, and calibration of the transformation between the measurement references of the two subsystems. Finally, a high-precision double-specular-surface calibrator is employed to optimize the initial calibration parameters. Three measurement errors are introduced to evaluate the 3D measurement accuracy of the calibrator. By minimizing the defined measurement errors, the optimal calibration parameters are calculated. Comparative experiments were performed to verify the application effects of the global optimization system calibration method. When the initial system calibration method was used to obtain the system parameters, and then the complete 3D shape of a gauge block was reconstructed, the root mean square error (R_RMSE) of the distance between the two surfaces of the gauge block was 164 μm. When the global optimization system calibration method was employed to determine the system parameters and reconstruct the complete 3D shape of the gauge block, the R_RMSE for the distance between the two surfaces was reduced to 34 μm. The global optimization system calibration method effectively improves the 3D shape measurement accuracy of double-specular-surface objects, providing a technical reference for enhancing the calibration precision of multi-sensor optical measurement systems under non-common-view conditions.