Image Feature Matching Based on SuperPoint and SuperGlue
This article first introduces the principles of SuperPoint and SuperGlue algorithms,and achieves a complete deep neural net-work image feature matching process by cascading the two together. Based on the designed deep neural network,four sets of multi-source images were selected for image feature matching,and compared with five classic feature matching algorithms. The experimental results show that the deep neural network designed in this paper has strong stability,high robustness,multiple and reasonably distribu-ted recognition feature points,and high accuracy in feature extraction,description,and matching of multi-source images.
万方数据
维普
