Aiming at the problem of inaccurate matching in wide baseline matching and three-dimensional scenes matching

a false match removing algorithm for smooth motion probability estimation is proposed. A single feature point statistical model is constructed in proposed algorithm

and this model is applied to the entire region of interest

assuming that the statistical value of the feature point obey the binomial distribution

construct the probability function of matching point and calculate the probability of correct matching

thus remove the mismatched outliers. Combined with grid accelerated matching

fast search for feature points by divided grid. It is verified that this algorithm has higher matching accuracy than other algorithms by experimental comparison.