felixchenfy / ros_detect_planes_from_depth_img

A python node to detect planes from depth image by using RANSAC algorithm. Input/Output from/to ROS topics.

Detect Planes from Depth Image

Demo of usage:

ROOT=$(rospack find ros_detect_planes_from_depth_img)
rosrun ros_detect_planes_from_depth_img run_server.py \
    --config_file $ROOT/config/plane_detector_config.yaml \
    --depth_topic test_data/depth \
    --color_topic test_data/color \
    --camera_info $ROOT/data/cam_params_realsense.json

The main scripts are: plane_detector.py, run_server.py, and config/plane_detector_config.yaml.

For a fast unit test, run: bash run_unit_test.sh

Demo:

Method:

(1) Create point cloud from depth image
(2) while RANSAC hasn't failed:
(3)    Use RANSAC to detect a plane from point cloud.
(4)    Remove the plane's points from point cloud.

Contents:

• 1. Results are published to ROS topics
• 2. Test without ROS
• 3. Test in ROS
  • 3.1. Publish color and depth images
  • 3.2. Run plane detection server
  • 3.3. View results in rviz

1. Results are published to ROS topics

$ rostopic list
/detect_plane/colored_mask
/detect_plane/image_viz
/detect_plane/results

The plane detection results are published to the topic detect_plane/results, whose message type is msg/PlanesResults.msg:

int32 N                 # Number of detected planes.

# In the following arrays, 
# the planes' parameters are concatinated one by one.

float32[] norms         # Nx3. Plane normal (nx, ny, nz).
float32[] center_3d     # Nx3. Plane 3D center (cx, cy, cz).
float32[] center_2d     # Nx2. Plane 2D center on the image (px, py),
                        #   which means {px}th column, and {py}th row.
uint16[] mask_color     # Nx3. Plane mask color (blue, green, red). 
                        # Each color's range is [0, 255]

2. Test without ROS

python plane_detector.py

3. Test in ROS

3.1. Publish color and depth images

Please download another package for publishing color and depth images:

cd ~/catkin_ws/src
git clone https://github.com/felixchenfy/ros_pub_and_sub_rgbd_and_cloud

Then, we can publish the data:

ROOT=$(rospack find ros_detect_planes_from_depth_img)
rosrun ros_pub_and_sub_rgbd_and_cloud pub_rgbd_and_cloud.py \
    --base_dir $ROOT \
    --config_file $ROOT/config/pub_rgbd_config.yaml

The settings are in config/pub_rgbd_images_config.yaml:

3.2. Run plane detection server

ROOT=$(rospack find ros_detect_planes_from_depth_img)
rosrun ros_detect_planes_from_depth_img run_server.py \
    --config_file $ROOT/config/plane_detector_config.yaml \
    --depth_topic test_data/depth \
    --color_topic test_data/color \
    --camera_info $ROOT/data/cam_params_realsense.json

3.3. View results in rviz

ROOT=$(rospack find ros_detect_planes_from_depth_img)
rosrun rviz rviz -d $ROOT/config/rviz_for_test_data.rviz