`voxel_min_points` parameter not working

[Srijan]

We observed "ghost points" from noise generated by depth cameras as seen in the images below. Ghost points are marked in blue, and it comes at random.

To mitigate this issue we looked into voxel_min_points and set it to different values ranging from 5 to 50. But, this didn't have any effect as seen in the image below -

Here is the plugin param config -

  enabled:               true
  voxel_decay:           20     #seconds if linear, e^n if exponential
  decay_model:           0      #0=linear, 1=exponential, -1=persistent
  voxel_size:            0.05   #meters
  track_unknown_space:   true   #default space is unknown
  observation_persistence: 0.0  #seconds
  max_obstacle_height:   1.0    #meters
  unknown_threshold:     15     #voxel height
  mark_threshold:        0      #voxel height
  update_footprint_enabled: true
  combination_method:    1      #1=max, 0=override
  obstacle_range:        3.0    #meters
  origin_z:              0.0    #meters
  publish_voxel_map:     true   # default off
  transform_tolerance:   0.2    # seconds
  mapping_mode:          false  # default off, saves map not for navigation
  map_save_duration:     60     #default 60s, how often to autosave
  observation_sources:   rgbd1_clear rgbd1_mark
  rgbd1_mark:
    data_type: PointCloud2
    topic:  camera/depth_registered/points
    marking: true
    clearing: false
    min_obstacle_height: 0.03     #default 0, meters
    max_obstacle_height: 1.0     #defaule 3, meters
    expected_update_rate: 0.0    #default 0, if not updating at this rate at least, remove from buffer
    observation_persistence: 0.0 #default 0, use all measurements taken during now-value, 0=latest 
    inf_is_valid: false          #default false, for laser scans
    clear_after_reading: true    #default false, clear the buffer after the layer gets readings from it
    filter: "voxel"              #default passthrough, apply "voxel", "passthrough", or no filter to sensor data, recommended to have at one filter on
    voxel_min_points: 10          #default 0, minimum points per voxel for voxel filter
  rgbd1_clear:
    enabled: true                #default true, can be toggled on/off with associated service call
    data_type: PointCloud2
    topic: camera/depth_registered/points
    marking: false
    clearing: true
    min_z: 0                   #default 0, meters
    max_z: 7.0                   #default 10, meters
    vertical_fov_angle: 0.7      #default 0.7, radians
    horizontal_fov_angle: 1.04   #default 1.04, radians
    decay_acceleration: 500       #default 0, 1/s^2. If laser scanner MUST be 0
    model_type: 0                #default 0 (depth camera). Use 1 for 3D Lidar

Kindly suggest a workaround for removing ghost points and why changing the voxel_min_points param doesn't work in our case

[Steve Macenski]

What version of this software are you using and is it actually released in that distribution?

I see these "plus sign" items in the costmap but that's not what I'm seeing in the visualized STVL grid. What's going on there?

[Srijan]

Hello Steve,
We are using Melodic-devel Ros1 Distribution and download the software from binaries.
The camera used is Intel Realsense D455 and whole plugin is run on a Jetson Xavier NX.

The plus signs are the ghost points which comes at random locations and there is no object corresponding to it in the real world. Also the STVL grid only makes a cube voxel for every plus sign which I am not sure the reason it happens.

I was just wondering can voxel_min_points parameter fix this plus sign issue by maybe not assigning a voxel to these ghost points?

[Steve Macenski]

Also the STVL grid only makes a cube voxel for every plus sign which I am not sure the reason it happens.

That doesn't seem related to STVL, I'm not sure what that relates to though. I've never seen anything like that before unless you've made your parameters really wacky.

Print before this line the value of the min voxel size https://github.com/SteveMacenski/spatio_temporal_voxel_layer/blob/melodic-devel/src/measurement_buffer.cpp#L164. If you get a non-zero number, then your gripe is with pcl. If you get a zero-value, I can look into it.

I was just wondering can voxel_min_points parameter fix this plus sign issue by maybe not assigning a voxel to these ghost points?

I don't know what / how you managed to get those plus signs (other costmap layers after this?) so I doubt that's in any way related.

[nickovaras]

This is an issue with PCL. The original Voxel Filter implementation used the original XYZ pointcloud message format and worked with the min_points parameter. With the transition to the PointCloud2 format, the parameter was left there exposed but it doesn't do anything. If it ever gets fixed in PCL, then it will automagically work with STVL (and would be very handy indeed).

[Srijan]

Print before this line the value of the min voxel size https://github.com/SteveMacenski/spatio_temporal_voxel_layer/blob/melodic-devel/src/measurement_buffer.cpp#L164. If you get a non-zero number, then your gripe is with pcl. If you get a zero-value, I can look into it.
Moreover, we have used the default pointcloud from the ros realsense wrapper and no other costmap layers are added in our implementation.

I tried printing, but the code doesn't seem to enter that block, it doesn't print any value (zero / non-zero), not even any static string written in that line.

Those plus signs, in my opinion are due to the lighting conditions in the room, maybe due to reflections.

I have tried using two different Realsense - Realsense D415 and D455,
Below are some errors observed with Realsense d455 -

As you can see in the image above, the reflection due to the sunlight causes a random voxel to appear in the map and it is marked onto the costmap with an inflation which should not happen in a general case.

And, ghost points appear at locations that aren't in the field of view of camera as you can see above.

I am just looking for voxel_min_points to somehow remove these unwanted voxels to appear or mark in the costmap altogether.