A polarization angle sensor was designed to gain the azimuth angle of a navigation system by learning from desert ants using the polarized pattern for navigation. To improve the measuring accuracy of the polarization angle, the factors influencing the angular errors were analyzed and the compensation method was researched. Based on the polarization principles, how to obtain the information of polarized azimuth angle from a sky polarization pattern was discussed, the main factors influencing the angular error measurement was analyzed, and a polarization angle error measuring model was built. According to the model, the solution to the polarization azimuth angle was deduced and the Least Square Method(LSM) of error compensation was presented to indirectly compensate the angular error by estimating the error parameters of the model. An experiment was performed eventually when the weather was fine. The experimental results show that the angular error is less than 0.17 degree (3σ), which indicates that the accuracy of the polarization angle is improved. The experiment demonstrates that the algorithm is feasible to compensate angular errors.