The invention provides a high-precision indoor positioning method based on error migration learning, and the method comprises the following steps: 1), enabling a mobile terminal A to scan all indoor wifi signals at divided grid points and to obtain a matrix V of signal intensity data and positioning data with a positioning error, and enabling a mobile terminal B to scan all wifi signals at m to-be-measured points and to obtain a matrix V-bar of signal intensity data and positioning data; 2), carrying out the transposition and normalization processing of the matrix V and the matrix V-bar; 3), learning the positioning error information of the mobile terminal A, and obtaining a base vector matrix X and a base vector coefficient matrix S; 4), enabling the matrix V-bar to be decomposed under the base vector coefficient matrix S, and obtaining a base vector coefficient matrix S-bar; 5), employing the matrix S and the matrix S-bar to carry out the error correction of a positioning result of the mobile terminal B at the to-be-measured point through a similarity comparison method based on vectors, and obtaining a corrected positioning result.