基于组合图像特征与分层节点搜索的回环检测方法

目的 文中通过提出一种新的回环解决方案，平衡回环检测系统的高准确率与高运行效率。方法 提出一种利用组合图像特征与分层节点搜索的新方法。首先，计算一种原始图像的下采样二值化全局特征和经过改进的ORB(oriented FAST and rotated BRIEF)局部特征，将其存入图像特征数据库。其次，引入一种分层节点搜索算法，在数据库中搜索与当前图像特征最相似的全局特征作为回环候选。最后，利用改进的ORB特征进行局部特征匹配，验证候选图像，确定回环检测结果。结果 使用该算法在3个不同的数据集上进行验证，测试中每次回环检测的平均处理时间仅需19 ms。结论 实验结果表明，该算法在运行效率、准确率、召回率等方面均达到了领域内的先进水平。

Loop Detection Method Based on Combined Image Features and Hierarchical Node Search

The work aims to propose a loop solution to balance the high precision and high efficiency of loop detection system. A new method based on combined image features and hierarchical nodes search algorithm was proposed. Firstly, a down-sampled binary global feature of the original image and improved ORB local feature were calculated and stored in the image feature database. Secondly, a hierarchical node search algorithm was introduced to search the database for the global feature most similar to the current image feature as a loopback candidate. Finally, the improved ORB features were applied to local feature matching to verify the candidate images and confirm the results of loop detection. The algorithm was validated on three different data sets, and the average time of each loop detection in the test was only 19 ms. The experimental results indicate that the algorithm has reached the advanced level in terms of operation efficiency, precision and recall.