针对 Oriented FAST and Rotated BRIEF (ORB)算法缺少尺度不变特性，误匹配率高，易造成图像拼接质量差等问题，本文提出一种用于图像拼接的改进ORB算法。首先使用一种基于尺度不变的特征检测算法对图像进行特征点检测，然后用ORB描述算法对特征点进行特征描述，用ORB匹配算法进行粗匹配，再用双向匹配和Random Sampling Coherence(RANSAC)算法对匹配点进行精匹配和提纯，进一步提高其正确率，最后使用渐入渐出加权融合完成图像拼接。实验结果表明，本文方法在图像的缩放和旋转、模糊、光照强度、拍摄视角等情况下都具有优秀的鲁棒性和稳定性，是一种实时性强、准确度高、拼接质量优秀的图像拼接方法。

In order to solve the problems that the Oriented FAST and Rotated BRIEF (ORB) algorithm does not have the feature of scale invariance, the highly false positive rate and the bad quality of the spliced image, this paper provides an improved ORB algorithm for image stitching. Firstly, a feature-based detection algorithm based on the same scale is used to detect the feature points of the image. Then the features of the detected feature points are described by using the ORB descriptor, and then the ORB matching algorithm is used for the rough matching. The bidirectional matching method and Random Sampling Coherence (RANSAC) algorithm are used to match and refine matching points to further improve its accuracy. Finally, the image mosaic is completed by fade-in-weight fusion. The experimental results show that the proposed algorithm is robust and stable against different illumination conditions, such as different rotation angles, low resolution and varying scale. It is a short time-consuming, high-precision and good image mosaic method.