Cylinder and Plane Extraction from Depth Cameras
Implementation of the method proposed in:
P. Proenca and Y. Gao, Fast Cylinder and Plane Extraction from Depth Cameras for Visual Odometry, IROS, 2018
https://arxiv.org/abs/1803.02380
Note: The parameters are fine-tuned for detecting large surfaces with Kinect 1 and Structure sensor for VO. For other applications, these may need to be modified.
- OpenCV
- Eigen3
RGBD Sequences w/ cylinders are available for testing: Here
Download a .zip file and unzip it into ./Data
Tested with Ubuntu 14.04 & 16.04
To compile, inside the directory ./CAPE
which contains the CmakeLists.txt, type:
mkdir build
cd build
cmake ..
make
To run the executable, use the following format:
./cape_offline <cell_size> <sequence_name>
where the first argument is the cell size in pixels (recommended 20)
and the second argument is a folder stored in:./Data
that contains the image and calibration files.
For example, if the target sequence is the 'tunnel' (assuming this was downloaded as specified above),run:
./cape_offline 20 tunnel
Tested configuration: Windows 8.1 with Visual Studio 12
This version includes already a VC11 project. Just make the necessary changes to link the project with OpenCV and Eigen.