Image feature registration is a critical step for stitching and generating large-field images during the inland navigation of ships. To address the problems of sparse water surface feature points and low efficiency in traditional feature matching algorithms for image registration in inland navigation environments, this paper proposes a feature matching method based on image super-resolution reconstruction. Firstly, the input images are subjected to super-resolution reconstruction using generative adversarial networks to enrich image details and increase the number of image feature points. Secondly, the ORB operator and BEBLID algorithm are employed for feature point detection and description. Then, coarse matching is performed based on Hamming distance. Finally, an improved random sampling consistency algorithm is utilized to further eliminate gross errors and purify inliers, achieving robust matching results. The study conducts experiments using five sets of inland navigation environment images with challenges such as low visibility, varying lighting conditions, scale changes, blur, and rotation. The results demonstrate that the proposed approach, leveraging image super-resolution reconstruction for feature point extraction, achieves an increased number of feature points and outperforms comparative algorithms in terms of matching accuracy and speed. This method meets the requirements of high-precision and real-time image matching in inland navigation environments.