Problem with ground filter about octomap_mapping HOT 8 OPEN

To be simple, you misunderstood the parameters.
If you look inside the code and you will find some functions in <octomap_server.cpp>.
in function OctomapServer::insertCloudCallback, author use function void OctomapServer::filterGroundPlane to do the filter work.
The base idea is to find a plane which is vertical to the axis-Z in world cordinate. zmin zmax are used to decide the inliers.
Enlarge the range between zmin~zmax will put too many points into account, which affect the ratio of inliers to outliers.In this situation, the efficiency of RANSAC decrease and pcl::SACSegmentation may not find the plane in 200 loops .
If your drone fly in different altitude， you should carefully modify your urdf file and activate your slam algorithm in the ground, so your drone can recognize the ground as z=0.After that,octomap_server should work properly.
Good luck!

from octomap_mapping.

That PCL error / warning appears when there can be no plane estimated, usually when there are to few or no points. Maybe the parameter sensor_model/max_range is too low?

You could also try debugging with rqt_console, depending on the log level set you'll see exactly what octomap_server is doing.

from octomap_mapping.

Hi ! I'm having the same issue, with or without defining a sensor_model/max_range

For example I run:

rosrun octomap_server octomap_server_node cloud_in:=/pcl_xyz _occupancy_min_z:=-.85 _occupancy_max_z:=0.5 _filter_ground:=true _ground_filter/plane_distance:=0.8

which includes points belonging to the ground.

To get similar errors:

[INFO] [1497343346.217816837]: PCL segmentation did not find any plane. [ WARN] [1497343346.217868956]: No ground plane found in scan [ERROR] [1497343346.426143499]: Transform error for ground plane filter: "base_footprint" passed to lookupTransform argument target_frame does not exist. , quitting callback. You need to set the base_frame_id or disable filter_ground. ^C[pcl::SampleConsensusModelPlane::optimizeModelCoefficients] Not enough inliers found to support a model (0)! Returning the same coefficients. [pcl::SampleConsensusModelPlane::optimizeModelCoefficients] Not enough inliers found to support a model (0)! Returning the same coefficients. [pcl::SampleConsensusModelPlane::optimizeModelCoefficients] Not enough inliers found to support a model (0)! Returning the same coefficients.

The PointCloud2 I'm using as input for octomap_server is published by a Parrot SlamDunk ROS node, it has a reasonable amount of points, although I admit not that much belonging to the ground, and thus, enabling a successful plane detection.

from octomap_mapping.

If there are too few points for a reasonable ground plane fit then you should not use the ground filter.

However, I'm seeing a Transform error in your log. Could it be that you're missing the proper base and base_footprint tf transforms?

from octomap_mapping.

Indeed ! I just tried it again, this time with proper parameters for octomap_server ~frame_id and ~base_frame_id (view_frames pdf return here), getting the no ground plane found warning again.

Command-line messages:

[pcl::SampleConsensusModelPlane::optimizeModelCoefficients] Not enough inliers found to support a model (0)! Returning the same coefficients. [ INFO] [1497517506.648828231]: PCL segmentation did not find any plane. [ WARN] [1497517506.648858067]: No ground plane found in scan [pcl::SampleConsensusModelPlane::optimizeModelCoefficients] Not enough inliers found to support a model (0)! Returning the same coefficients.

Command-line command:

rosrun octomap_server octomap_server_node cloud_in:=/pcl_xyz _occupancy_min_z:=-1 _occupancy_max_z:=0.5 _frame_id:=/map _base_frame_id:=/cam0 _filter_ground:=true _ground_filter/plane_distance:=0.7

I have quite a lot of points in my PointCloud2 (.pcd example file here), though not on the ground, I don't know why my sensor doesn't detect much there...

Anyway, if I can't do anything with octomap_server parameters to force the ground filter to work I guess I'll just have to find another way to do it. Thanks for your answer ! 👍

from octomap_mapping.

I tried octomap output with my own pcl ground filter node. it works fine. So you can use it same way as well

from octomap_mapping.

Did anyone find a solution?

from octomap_mapping.

Did anyone find a solution to this ? Which parameter should i tweak in the ros node to solve this ?
Is it pointcloud_max_z and pointcloud_min_z ? What range should I use.
I did not really understand the last answer

from octomap_mapping.