基于全局特征与局部特征的图像分级匹配算法研究及应用

摘要

Aiming at the problems that the image feature extraction and match algorithm should have high accuracy,good performance in real-time and robustness when the mobile robot navigated based on topological map,the paper proposed a new image hierarchical matching algorithm based on the global feature and local fusion.First,the improved shape context algorithm was applied to the in-put images to be matched to extract the global features which were matched roughly with the images in the database,obtaining 3 images that had the highest similarity to the current image which consti-tuted a temporary image database.Then,the improved SIFT algorithm were used to extract local fea-ture of the input images which were matched accurately with the local feature of the 3 images in the temporary image database,finally obtaining the image with highest similarity as the match results. The image hierarchical matching algorithm based on the global feature and local feature combined the improved shape context algorithm based on global features and the improved SIFT algorithm based on local features,so as to achieve the purpose of complementary advantages.The experimental results show that this algorithm can effectively improve the matching efficiency and reduce the running time when the mobile robots navigate based on topological map.%移动机器人基于拓扑地图导航时要求图像特征提取与匹配算法具有高的精度和鲁棒性、良好的实时性，针对此，提出了基于全局特征和局部特征的图像分级匹配算法。首先对输入的待匹配图像应用改进的形状上下文算法提取全局特征与图像库中图像进行遍历粗匹配，得到与当前待匹配图像相似度最高的3幅图像并构建临时图像库；然后利用改进的 SIFT 算法提取输入图像局部特征与临时图像库中3幅图像的局部特征进行精确匹配，最终得到与待匹配图像相似度最高的图像作为匹配结果输出。所提出的图像分级匹配算法将基于全局特征的改进形状上下文算法和基于局部特征的改进 SIFT算法相结合，从而达到优势互补的目的。实验结果表明，该算法在机器人基于拓扑地图导航过程中有效地提高了图像匹配效率，缩短了运行时间。