At present

least squares algorithm is generally used to calibrate cameras. Constraint condition selections have a great effect on estimating precision. On the basis of analysis to former selections of constraint condition

a new constraint condition selection method is rpesented. Computing formulas of parameters are got after mathematical derivation. The algorithm avoids solving nonlinear equation. After constraint relations between parameters are observed and studied

a constraint condition is selected using the property that orthogonal matrix ensures Frobenius norm to be unvarying. Errors are distributed in the light of constraint relations between parameters

so precision of camera calibration can be improved.