In speckle vision measurement, markers are usually used to improve the measurement efficiency. To overcome long matching time and low matching accuracy in traditional marker matching process, we propose a new method for marker matching by using improved Kanade-Lucas-Tomasi (KLT) algorithm. Firstly, we measure the marker to establish initial matching point based on the improved speeded up robust feature (SURF) algorithm. Secondly, we use the improved KLT algorithm to achieve the marker matching. Thirdly, we use the constraint condition based on max bidirectional error to delete the mismatched points and improve the reliability of maker matching. Finally, we make a matched experiment to verify the marker coated on the speckle region of the wing model during the wing flutter measuring. The results show that, compared with traditional scale-invariant feature transform (SIFT) and SURF matching methods, the proposed method reduces the matching time by 75.9% and 42.8%, and improves the matching accuracy by 30.6% and 22.2%, respectively.