ORB 特征点 (Oriented FAST and Rotated BRIEF)

Posted

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了ORB 特征点 (Oriented FAST and Rotated BRIEF)相关的知识,希望对你有一定的参考价值。

ORB 是一种具有局部不变形的特征点特征,是FAST特征点和BRIEF描述子的结合。

FAST特征:

1.从图片中选择像素点P,亮度L

2.设定threshold T

3.consider a circle with R=3 , meaning 16 points in total

4. if there are n consective points on the circle, > I+T or <I-T , it is a corner point. n can be 12 or 9

 

we can use high speed test to fastly eliminate large unsuitable areas:

we check points No.1 9 5 and 13. if 1 and 9  meet requirement 4, then check 5 and 13. if it is a corner point ,there 3/4 of the circle points will meet 4. if not, it is not a corner point.

Note: we cannot use HST if n<12

BRIEF descriptor:

1. find a patch around a feature point

2. find Nd point pairs inside the patch

3. for every point pairs, if I(1)>I(2) , its corresponding binary value=0 , else 1

4.we can thus generate a binary string with length=Nd

5. we can then use hammering distance to describe these descriptors.

以上是关于ORB 特征点 (Oriented FAST and Rotated BRIEF)的主要内容,如果未能解决你的问题,请参考以下文章

OpenCV中ORB特征提取与匹配

SIFT,SURF,ORB,FAST,BRISK 特征提取算法比较

opencv ORB特征匹配

ORB_SLAM2 源码解析 ORB特征提取

《视觉SLAM十四讲》课后习题—ch7(更新中……)

ORB算法与opencv实现