Pointcloud Evaluation Tool

---

Computes the Mean Map Entropy and the Mean Plane Variance of a point cloud.

Furthermore a pointcloud is generated that encodes the Map Entropy and Plane Variance in each point.

Filtering by field type was also included. ROS-mode and ROS-independent mode support was also added.

Credits

This code is based on the work by David Droeschel & Jan Razlaw:

Razlaw, J., Droeschel, D., Holz, D., & Behnke, S. "Evaluation of registration methods for sparse 3D laser scans." 
In 2015 IEEE European Conference on Mobile Robots (ECMR), September 2015.

Here are the links to their Github and their Paper

Install

---

ROS mode

The following will also produce the ROS-independent build so you can run the program directly.

mkdir <ROS-WORKSPACE>/mean_map_entropy/src -p
cd <ROS-WORKSPACE>/mean_map_entropy/src
git clone <THIS-CODE-REPOSITORY>
cd ../
catkin build  # <- you can also use: colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release

ROS independent mode

mkdir <SOME-FOLDER>/mean_map_entropy -p
cd <SOME-FOLDER>/mean_map_entropy
git clone <THIS-CODE-REPOSITORY>
mkdir build
cd build
cmake ..
make

Usage

---

ROS mode

cd <ROS-WORKSPACE>/mean_map_entropy
source install/setup.bash
roslaunch mean_map_entropy mean_map_entropy.launch input_topic:=<INPUT_CLOUD_TOPIC> output_topic:=<OUTPUT_CLOUD_TOPIC> output_frame:=<FRAME_NAME> 

You can use other arguments as defined in the launch file. And you can use RVIZ to visualize the output pointcloud topic.

ROS independent mode

./mean_map_entropy_pcl path/to/pointcloud.pcd [-stepsize int] [-radius double] [-punishSolitaryPoints] [-planevariance] [-ascii] [-minNeighbors int] [-passfilter] [-filterField string] [-minLimit double] [-maxLimit double]

Available arguments are as follows:

• stepsize: stepsize used iterating through the pointcloud (default: 1)
• radius: radius used for search for neighboring points (default: 0.3)
• punishSolitaryPoints: (flag) punishes points with bad values where the number of neighbors is under minNeighbors
• planevariance: (flag) whether to compute the plane Variance or not
• ascii: (flag) whether to save the output point cloud file as ASCII PCD type
• minNeighbors: threshold (default: 15)
• passfilter: perform band pass filter on the values of the specified field
• filterField: the field name for filtering
• minLimit: minimum limit for filtering
• minLimit: maximum limit for filtering

Use the point cloud viewer of your choice to visualize the output (e.g. pcl_viewer).

TODO

• Add ROS support
• There is a bug when generating the output pointcloud which removes the intensity field and adds extra annonymous ('-') fields
• Computing map entropy and plane variance from a grid was started but the code is not finished
• Other forms pointcloud evaluations, such as the mean shift distance
• It will be nice to add more filtering options, like filtering by multiple fields and with some regular expressions (ex. intensity in [min, max] AND entropy in [min, max])