The invention relates to error correction of a vector sensor, in particular to a calibration compensation method for a triaxial vector sensor and a biaxial vector sensor, and solves the problems that the calibration compensation method of the existing multi-axial vector sensor fails to take all factors causing measurement error into account and the like. In the method, an error correction mathematical model of the vector sensor is built according to the relation of actually measured output Sm and theoretical output Se of the vector sensor, i.e. Sm=KSe + S0 and according to an error coefficient matrix K=K1K2, wherein, K1 and K2 are respectively a sensitivity error coefficient matrix of the triaxial vector sensor and a non-orthogonal error coefficient matrix among three measured axes; and an effective method is utilized to obtain correct zero-offset S0 and a correction coefficient matrix K-1 in the error correction mathematical model, and finally the error correction mathematical model of the measured vector sensor is obtained to carry out calibration compensation on the measuring result of the vector sensor. The method achieves calibration compensation taking into account of all the factors causing the measurement error of the multi-axial sensor, thus improving the accuracy of measuring results; and the method has concise and convenient process and no excessively high requirements on hardware equipment, thus being suitable for various vector sensors.