Application of improved SURF algorithm in real scene matching and recognition

Aiming at the problems of insufficient extraction of stable features by SURF algorithm, high matching error rate, and loss of correct matching information during matching purification, this paper improves from two aspects: feature extraction and feature matching. First multi-scale FAST corners is constructed to make the extracted corners scale-invariant; second, use the improved CANNY algorithm to increase the number of stable features according to the double threshold adaptively set by the Maximum Inter-Class Variance Algorithm (OTSU); then the SURF feature descriptor is constructed; finally, the feature matching effect is improved by combining KNN bidirectional matching and RANSAC algorithm. The experimental results show that the improved algorithm has better results than the original algorithm in both the number of extracted feature points, stability, and final matching accuracy.