Research on robot scene recognition based on improved feature point matching algorithm

A new feature description method based on the fusion of fast retina keypoint (FREAK) and rotation-aware binary robust independent elementary features (rRBRIEF) is proposed to realize effective combination efficiency accuracy two descriptions. In addition, in elimination stage mismatched point pairs, by setting base its neighborhood, an improved neighborhood parallel random sample consensus (RANSAC) algorithm achieve efficient operation multiple local neighborhoods. The matching existing were tested different scales, rotations, illuminations, fuzzy data sets. experimental results show that improves average scene recognition 18.21%, 15.58%, shows good robustness.