利用计算机视觉对桥梁结构进行位移测量，常见的方法有特征点算法和光流法。为解决光照变化对光流法的精度影响，以及特征点法难以对目标进行实时跟踪等问题，以定向FAST与旋转BRIEF（oriented FAST and rotated BRIEF, ORB）特征点算法提取目标特征点，并结合LK（Lucas-Kanade）光流法进行跟踪。通过两种算法的互补融合，并设定周期性更新策略，有效减少误差积累。在室内进行了钢板梁模型激振试验，探讨光照变化对算法性能的影响。将算法得到的位移数据与激光位移计测量结果进行对比，以验证其准确性。此外，还将该算法与FAST-LK、SIFT-LK、Shi-Tomasi-LK、siamRPN++等算法进行比对分析。结果表明：该算法得到的位移曲线拟合程度在95%以上，误差控制在5%之内，验证了其可行性，同时，试验表明该方法对光照变化具有一定的鲁棒性，并能达到实时性的要求。

The common methods for displacement measurement of bridge structures using computer vision include feature point algorithm and optical flow method. To address the influence of illumination variation on the accuracy of the optical flow method and the difficulty of the feature point method in real-time tracking of the target, the target feature points were extracted using the oriented FAST and rotated BRIEF (ORB) feature point algorithm, and tracking was carried out in combination with the Lucas-Kanade(LK) optical flow method. By complementing and integrating the two algorithms and setting a periodic update strategy, the accumulation of errors can be effectively reduced. The vibration test of the steel plate beam model was conducted indoors to explore the influence of light variation on the performance of the algorithm. The displacement data obtained by the algorithm is compared with the measurement results of the laser displacement meter to verify its accuracy. In addition, the algorithm was compared and analyzed with algorithms such as FAST-LK, SIFT-LK, Shi-Tomasi-LK, and siamRPN + +. The results show that the displacement curve fitting degree obtained by this algorithm is above 95%, and the error is controlled within 5%, verifying its feasibility. Meanwhile, the experiments indicate that this method has certain robustness to illumination changes and can meet the requirements of real-time performance.