To tackle the issues of slow searching and insufficient precision in feature detection and matching using the SURF algorithm, an improved image matching algorithm based on SURF algorithm has been proposed. Compared with SURF algorithm, by optimizing key components, this novel approach makes the feature detection and matching faster and more accurately. First, the conventional SURF algorithm is used to find the target image's feature points, and generate a 64-dimensional vector to represent the feature description of each point. Then, several nearest neighbors of each feature point are quickly found by building a KD- Tree. In order to optimize the feature points in both directions, the RANSAC algorithm is utilized to remove the set of incorrectly matched feature points. The findings of the experiments indicate that the improved algorithm suggested in this research enhances matching precision and speeding up feature point matching.