Aiming at the problems of long time-consuming and insufficient matching rate of feature point matching in current image matching methods

an image matching detection method based on improved ORB and grid-based motion statistics is proposed. The SURF algorithm is used to extract the image feature points

the four-layer image pyramid is constructed to assign the scale invariant features to the feature points

and the direction information is added by the gray-scale centroid method. Smoothness of the motion results

in the correctly matched feature points have more matching points in the neighborhood. After the rough matching is achieved

the feature registration algorithm of grid-based motion statistics is used to complete the screening and culling of feature point mismatching. The experimental results show that the registration precision of the proposed algorithm is greater than 94 % when the image is rotated

scaled

blurred and brightness changed

it has high matching recognition rate and robustness.