【摘要】 在图像拼接过程中，使用固定阈值的旋转不变性二进制描述（Oriented fast and rotated brief,ORB）算法检测出的特征点在特征匹配阶段会产生较多误匹配，从而导致拼接的图像在拼接缝处产生重影。针对此问题，提出一种基于改进ORB的图像拼接算法。首先，使用自适应算法将ORB的固定阈值替换为动态阈值对特征点进行检测；然后，使用K近邻（K-nearest neighbor,KNN）算法进行特征点粗匹配，再用随机抽样一致性（Random sampling consensus,RANSAC）算法对特征点进行精匹配；最后，通过最佳缝合线法和渐入渐出法对图像进行拼接。实验结果表明，相较于传统的ORB算法，利用所提出的算法时所需的特征点检测时间、匹配时间明显减少，匹配正确率明显提高，拼接缝重影被有效消除。更多还原

【Abstract】 During image mosaicing, the feature points detected by the fixed threshold oriented fast and rotated brief(ORB) algorithm will produce more mismatches in the feature matching stage, which results in a double image of the stitched image at the seam. To solve this problem, an image mosaic algorithm based on improved ORB is proposed. Firstly, the fixed threshold of ORB is replaced by the dynamic threshold to detect the feature points by using adaptive algorithm, and then the K-nearest neighbour(KNN)algorithm is used to perform the rough matching of the feature points, then, random sampling consensus(RANSAC) algorithm is used to match the RANSAC points, and the image is stitched by the best suture method and the gradual-out method. The experimental results show that compared with the traditional ORB algorithm, the proposed algorithm head less feature point the detection time and matching time, and improve the matching accuracy with eliminating the stitched ghosting effectively.更多还原