Improved Correspondence Point Search and Image Matching Accuracy for Disturbanced Images Using TD(0) Method

抄録

This paper presents a novel reinforcement learning approach to enhance image matching, whereby the corresponding point candidates are detected from the feature points extracted from each of the two images. In general, robust estimation methods such as random sample consensus (RANSAC) are used to select valid corresponding points from the candidates. Therefore, we addressed the limitations of RANSAC random selection in the image-matching process, evaluating various reinforcement learning strategies, including deterministic and probabilistic approaches, and different value update mechanisms. The findings indicated that a probabilistic approach with suitable value updates provides a more robust solution for space-based navigation systems.