Dense 3D Reconstruction from Stereo in ROS

This is a ROS package for real-time 3D reconstruction from stereo images using either LIBELAS or stereo block matching from OpenCV , for generating dense disparity maps. The 3D point cloud is also generated and can be visualized on rviz.

• Author: Pushyami Kaveti

Dependencies

• ROS Kinetic
• cmake
• gflags
• Boost
• OpenCV == 3.3.1

Stereo Calibration

The intrinsic and extrinsic calibration parameters of a pair of cameras should be stored in a .yml file. A stereo pair can be calibrated using a checkerboard rig using openCV or Kalibr. A sample calibration file is provided in the calibration folder

Compiling

Clone the repository of the ROS package in src/ folder of a catkin workspace :

$cd catkin_ws/src
$ git clone  https://github.com/PushyamiKaveti/stereo_dense_reconstruction.git

and execute

$cd catkin_ws 
$catkin_make 

Running the stereo app

$ ./bin/stereo_app [OPTION...]

Usage: dense_reconstruction [OPTION...]
   --left_topic=STR       Left image topic name
   --right_topic=STR      Right image topic name
   --calib_file=STR       Stereo calibration file name
   --calib_width=NUM      Calibration image width
   --calib_height=NUM     Calibration image height
   --is_ros=bool          is the dense reconstruction info published to ROS or saved to disk.
   --algo=NUM             which stereo algorithm to use ELAS=1, SGBM=2    
   --output_dir=STR       directory where the rgb and depth images should be saved      
   --color                for color or mono images
   --debug=bool           Set true for debug mode

if is_ros flag is true dense disparity grayscale image is published on the topic /camera/left/disparity_map and the corresponding point cloud on the topic /camera/left/point_cloud. Otherwise the depth map is saved onto disk under output_dir/depth folder. The depth map written to disk is not the disparity, but the actual depth image scaled by 5000.

A sample rosbag and the corresponding calibration file can be found here

License

This software is released under the GNU GPL v3 license.