Spatial failure satellite loopback detection method based on improved RANSAC verification

The invention discloses a space failure satellite loopback detection method based on improved RANSAC verification, and belongs to the field of visual navigation. The method mainly comprises the following steps of: extracting three features of feature points, local feature vectors and global feature vectors of an image through an open source deep learning method HF-Net; secondly, using an HNSW quick retrieval method to insert the global feature vector of the current image into an HNSW graph database, calculating the similarity between the global feature vector of the current image and the features in the database, obtaining an image index with the highest similarity, and forming a loopback candidate frame set; and finally, the local features are used for image matching verification, the optimal loopback image is determined, the accuracy and stability of loopback detection are improved, and the application prospect is wide. 本发明公开一种基于改进RANSAC校验的空间失效卫星回环检测方法，属于视觉导航领域。所述方法主要包括以下步骤：开源深度学习方法HF-Net，提取图像的特征点、局部特征向量以及全