关键词: 无人机视觉导航, 特征提取, 图像匹配, 混合注意力, 特征强化

Abstract: In order to improve the accuracy of image matching in UAV visual navigation, a UAV image matching model combining hybrid attention mechanism and feature reinforcement is proposed. Firstly, a visual geometry network feature extractor based on attention mechanism was designed to effectively associate similar features of images; In addition, a Transformer model that integrates feature enhancement networks was proposed in the process of feature point matching, which enhanced the feature information and improved the accuracy and robustness of image matching. The proposed model is trained and tested on the public data set. The experimental results show that compared with the matching model, the accuracy rate and recall rate of the model are increased by 1.85% and 0.22% respectively, which fully demonstrates the superiority of the model.

Key words: UAV visual navigation, feature extraction, image matching, hybrid attention, feature enhancement