An Iterative Selection Matching Algorithm Based on Fast Sample Consistency

The matching algorithm of feature point pairs is a research hotspot in machine vision at present. However, in the current mainstream algorithm, the method to deal with the mismatched point pairs generated in the matching process depends on the distance between descriptors as the constraint condition. Although the mismatching can be eliminated, the number of correctly matched feature points will also decrease. Therefore, an iterative selection algorithm based on fast sample consistency is proposed. First, the Fast Sample Consensus (FSC) is combined with the iterative selection matching algorithm to find the maximum consistent set; Then, the minimum optimal sampling number is determined according to the distance of the matching point pair, and four groups of data are selected to calculate the transformation model; Finally, the transformation model is applied to the image to be registered. Experiments show that the algorithm increases matching accuracy by almost 8% while simultaneously saving roughly 25% of the time cost.