Using rtabmap as backend to colorize points

[dinossht]

I am currently using rtabmap_slam node with fastlio as odometry. I want to know if I can use rtabmap_slam node to just accumulate odometry and point loud data from fastlio node and construct map, as well as assembling camera images as for coloring purposes later. My question is whether it possible to reduce processing heavy activities such as loop closure detection, icp matching etc in rtabmap. I experience that fastlio already provides accurate odometry and pointcloud, so it seems unnecessary to redo map optimization etc. I want to basically use rtabmap to merge fastlio odometry and pointcloud with rgbd images to output colored pointcloud and map. Is this possible?

[matlabbe]

Yes, it is possible.

My answer would be similar to scenario one of this post:

# Values should be all "string" type 
'Kp/MaxFeatures': '-1',              # Disable visual loop closure detection
'RGBD/ProximityBySpace': 'false',    # Disable proximity detection
'RGBD/CreateOccupancyGrid': 'false', # Disable occupancy grid
'RGBD/LinearUpdate': '0.1',          # Adjust minimum motion to add new nodes to the map
'RGBD/AngularUpdate': '0.1',         # Adjust minimum motion to add new nodes to the map
'Rtabmap/DetectionRate': '1',        # Adjust at which rate new nodes are added to the map

Note that you could still use rtabmap-databaseViewer to add manually loop closures afterwards if you find that fastlio drifted too much.

[dinossht]

Thank you!