wolfv/osm_pc_mapping

Might map OSM outlines to pointclouds in the future

OSM_PC_Mapping

Might map OSM outlines to pointclouds in the future

###Bundler Parser### Compile:

catkin build bundler_parser

Run:

Camera Query:

devel/lib/bundler_parser/parse_bundler --alsologtostderr --colorlogtostderr --file=[full_path]/file_name.out --query=camera --indices="1 3 12"

3D Point Query:

devel/lib/bundler_parser/parse_bundler --alsologtostderr --colorlogtostderr --file=[full_path]/file_name.out --query=3dpoint --indices="1 5 8"

###Point Cloud Segmentation### Compile:

catkin build pc_segmentation

Run:

Segment Point Cloud:

devel/lib/pc_segmentation/segment_pc --alsologtostderr --colorlogtostderr --bundler_file=[full_path]/file_name.out --outline_file_path=[storage_path] --show_cameras=true --show_cloud=true --show_normals=false --segmentation_threshold=0.001 --search_radius=0.4

Point Cloud matching

Dependencies:

numpy
python-pcl
opencv-python
scipy
matplotlib

Obtaining Ground truth

Execute osm/visualization.py

Filtering

In folder ransac, cmake ., make and then executable filters will yield filtered output from test_pcd.pcd and save it as filtered_output.pcd.

Corners are found by using corner_extract.py on filtered output.

Then corners can be matched by using match_corners.py (note long runtime since dumb bruteforce method).

ICP

In libpointmatcher folder, executables are placed in libpointmatcher/examples.

Execute ./pmicp --config config.yaml --initTranslation 70,100 ../data/cl.csv ../data/cl2.csv to generate a mapping based on a Point-to-Point ICP.

The results can be inspected by opening ParaView and looking at the resulting visttep-files.