Location modes on mobile phones

[yb]

Hello, thank you for your work. I have generated a map using an rgbd camera (Realsence D435i). Now, I want to use this map for positioning on my mobile phone (the phone does not have the ROS environment or depth information). How should I do it? Are there any relevant tutorials? Basically, I just need to input the image from the mobile phone camera and obtain the current position and orientation. I've read other posts. Is the process of positioning actually loop closure detection? Advance thanks.

[matlabbe]

See related post here: http://official-rtab-map-forum.206.s1.nabble.com/Localize-in-a-pre-loaded-local-map-created-by-RGB-D-mapping-tp2383p2399.html

If you record a video from your phone, the approach above may work with RTAB-Map standalone app. You may need to calibrate your phone's camera like this.

Note however that mode won't compute visual odometry, it will just do re-localization, so it may not estimate a pose for every frame if it cannot estimate enough visual correspondences between the current frame and those in the map. Re-localization is done with same code than loop closure detection.

cheers,
Mathieu

[yb]

Thank you for your reply. Does calibrating the camera mean obtaining the camera parameters? Do I need to perform this calibration on all the cameras on my phone to obtain the correct pose? My understanding is that since the map is created using cameras with different parameters, new camera parameters are needed if relocalization is performed using different cameras.

[matlabbe]

Camera intrinsics (minimally fx,fy,cx,cy) are required to know the focal distance of the camera you are using for localization. These values could be already available through your phone's camera API (either iOS or Android). There could be an app showing all parameters.

Note that for a first test, you could just guess the intrinsics, though the pose's scale may be wrong.