ORB_SLAM代码阅读_注释整理

ORB特征点检测、匹配
1、http://www.javashuo.com/article/p-cqvdhkyp-np.html
2、特征点提取原理
3、ORB_SLAM特征点部分注释：
https://blog.csdn.net/u014797790/article/details/79776339
https://blog.csdn.net/weixin_43828675/article/details/113792110
ORB-SLAM2特点匹配代码分析
ORB描述符匹配源码解释
4、SLAM相关名词解释
https://blog.csdn.net/Jinxiaoyu886/article/details/88836762
5、DBOW2
https://blog.csdn.net/weixin_40893939/article/details/105715243
BOW 词袋
https://zhuanlan.zhihu.com/p/64221963
BOW特征匹配　
https://blog.csdn.net/KYJL888/article/details/102588059
6、sim3相似性变换求解
https://blog.csdn.net/moyu123456789/article/details/91947539#1.sim3%E7%9A%84%E7%AE%80%E5%8D%95%E6%A6%82%E8%BF%B0
7、ORB＿SLAM3单目源码梳理
参考博客
8、ORB-SLAM3 细读单目初始化过程（总结很好）
https://blog.csdn.net/shanpenghui/article/details/110003959
个人总结的一些概念：
Relocalisation：Relocalisation consists in locating the robot on an existing map when its pose is unknown. Relocalisation is indispensable if we want to reuse a previous map, otherwise the map becomes useless apart from localising the camera during a certain operation. In addition relocalisation is essential for recovering after a tracking failure, very common in visual SLAM due to occlusions (a person moving in front of the camera), untextured areas or fast camera movements.（来源论文：Fast Relocalisation and Loop Closing in Keyframe-Based SLAM）
Loop closing：Loop closing is the task of detecting when a robot is revisiting a previously mapped area, in order to correct the error accumulated in the robot trajectory during exploration. Tracking data association is able to close small loops. In the case of large loops, the prediction of the camera location accumulates so much error, that data association fails to establish correspondences between the current image and the map. The result is that the system behaves as if the image were showing a new place and thus duplicating previously
mapped areas.（来源论文：Fast Relocalisation and Loop Closing in Keyframe-Based SLAM）
tf-idf：term frequencyinverse document frequency，在文献检索中可以理解为词频，某个词语在文章中出现的频率。浅析 Bag of Feature