The invention discloses a rapid robust model estimation method based on adaptive ordering, and aims at solving a technical problem that a conventional rapid robust model estimation method is poor in practicality. The technical scheme of the invention is that the method comprises the steps: carrying out the initial ordering of image feature point pairs U before the multi-sampling and model estimation, and estimating an optimal model through multiple tests; obtaining a sampling subset Un which is formed by the former n feature point pairs with the highest quality in the image feature point pairs U in each test, and obtaining sampling samples from Un; carrying out the estimation of the model of a current test after the obtaining of the sampling sample of each test; stopping one of the tests only when the current iteration meets a stopping condition; recalculating model parameters through elements in an optimal consistent set, and finally obtaining an optimal model. Because a random sampling consistent model estimation algorithm based on adaptive ordering is employed, the method can improve the efficiency while maintaining the error matching eliminating effect, achieves the quick obtaining of a robust model estimation result, and is good in practicality.