什么是词汇树以及如何构建词汇树？
当涉及到大型数据库中类似图像或文本的快速查询时，我阅读了很多有关词汇树的内容。但我找不到任何关于什么是这样的词汇树以及如何构建其中一项功能的好的（易于理解的）描述。
I'm reading a lot about vocabulary trees when it comes to fast queries for similar images or texts in big databases. But I couldn't find any good (easy to understand) description on what such a vocabulary tree is and how to build one of features.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
词汇树是一种进行图像检索的紧凑方法。实现该算法基本上分为三个步骤，并且它高度依赖于其他计算机视觉技术，例如 SIFT 特征。
第一步是使用 sift 描述符构建 kmeans 树。该树的叶节点包含一“袋”的 sift 描述符。第二步是使用第一步中构建的词汇树构建图像数据库。您可以将此过程视为将图像量化到向量空间中。然后第三步是根据图像数据库查询图像。当然还有一些细节技巧比如倒排列表等。
这里有一个很好的词汇树实现 - libvot。它基本上遵循我上面描述的三个步骤。它使用 C++11 标准多线程库来加速构建过程，因此运行速度相当快。
这是关于它的原始研究论文 。近年来，它对计算机视觉界产生了很大的影响。
Vocabulary tree is a compact way to do image retrieval. There are basically three steps to implement this algorithm, and it highly depends on other computer vision techniques e.g. SIFT features.
The first step is to build a kmeans tree using sift descriptors. The leaf nodes of this tree contain a "bag" of sift descriptors. The second step is to build a image database using the vocabulary tree you build in the first step. You can view this process as quantizing an image into a vector space. Then the third step is to query the image against the image database. Of course there are some detailed techniques such as inverted list, etc.
Here is a good implementation of vocabulary tree - libvot. It basically follow the three steps I described above. It use the C++11 standard multi-thread library to accelerate the build process so it runs pretty fast.
Here is the original research paper about it. It has high impact on computer vision community these years.
词汇树是通过将视觉单词聚类成簇而形成的。为此应用了树结构，使其成为词汇树。下图会让事情变得清楚。
另请检查此：
Vocabulary tree is something which is made by clustering visual words into clusters. Tree structure is applied for this which makes it vocabulary tree. Following diagram will make things clear.
Also check this: