In order to achieve relative orientation and scale values between environmental information and reference information for path integration of UCAV cognitive navigation, a method based on iterative least-squares is proposed to solve high-precision similarity transformation parameters. SURF algorithm is used to extract high robustness feature points, then ratio method is taken to purify the matching pairs to get a point-to-point set, the elements in the set are further transformed into vectors, similarity transformation parameters are gained by iterative least-squares algorithm, then the perceived image's orientation is rotated backward, and the scale is adjusted inversely according to the gained parameters, circulate the operations until relative orientation and scale value are finally obtained. The simulation results show that the use of the above method can get high-precision orientation and scale parameters and anti-noise performance is superior to the least squares algorithm.