I am aware that algorithm for monocular SLAM is complex compared to stereo SLAM. But my question is if by any means it is possible to do SLAM using one camera why one should use two cameras to do the same thing?
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
6
The most important point is the scale. If you do monocular SLAM, your map will only be accurate up to scale so that you e.g. cannot compute the length of the travelled path in meters. The scale between your map and the world is not even constant over time so that if you come back to your starting point, it's going to be difficult to match the beginning and end of you map as they have a different scale.
-
$\begingroup$ Thanks for your answer @FooBar. But I believe that this limitation can by solved by using loop closure. Dont you think so ? $\endgroup$– vackyFeb 12, 2017 at 18:21
-
$\begingroup$ No. You can move your camera slowly over a toy city or attach it to a drone. Just by looking at the images, you are not able (no one is able!) to detect the difference. You can only add other sensors (e.g. pressure sensor from the drone) that give you the absolute scale. $\endgroup$ Feb 12, 2017 at 18:50
-
-
1$\begingroup$ If you're constrained to only one camera, it would probably make sense to use multiple frames to initialize the scale, as PTAM does: or another (fancier) way is what's called image alignment to obtain transformations: as in LSD-SLAM. $\endgroup$ Feb 13, 2017 at 3:45
-
2$\begingroup$ This can only help in fixing the scale, but never to get the correct metric scale. $\endgroup$ Feb 13, 2017 at 9:50