Skip to content

searchwingCV/droneCamBoatDetector

Repository files navigation

DEPRECATED CODE

For more info visit

wiki: https://wiki.searchwing.org/en/home/Development/Payload-Camera/searchwing-bilderkennung

ComputerVision for searchwing drones

alt text

  • Left: 3D View
    • "Cam": 3D-Position and orientation of drone camera
    • Pink points: the visible area in the image
    • Coordinates: possible tracked boats with their ID
  • Right: 2D View
    • Yellow boxes: Regions of Interest(ROIs) in the current image which will be associated to existing trackings
    • Red boxes: ROIs in the image which are too big
    • Green crosses: Tracked positions (with ID) which got recognized over several frames and therefore considered as boats
    • Blue crosses: Tracked positions which were created in previous frames

Install dependencies

Python 2.7

Install Opencv3 for python 2.7:

python -m pip install opencv-python

ROS

ROS handles the communication between the different modules, transforms 3d-data and provides a handy 3d visualization for development

Build/integrate src to ros framework

cd /catkin_ws/src/
https://github.com/searchwingCV/droneCamBoatDetector.git
  • Compile/install
cd /catkin_ws  
catkin_make install
source /catkin_ws/devel/setup.bash

Run on the bodensee-dataset

  • New Terminal: Start roscore
roscore
  • New Terminal: Start dataset playbag.
cd datasetDownloadPath
rosbag play bodenSee2018HalfRes.bag --clock --start=0 --rate=2
  • New Terminal: Setup environment:
source /catkin_ws/devel/setup.bash

Start boatdetector:

roslaunch searchwing bodenseeDataset.launch

Now you should see the detections in rviz like in the screenshot above.

If you want to edit/start/debug the boadDetectorNode.py with a editor, please ensure you run

rosparam set use_sim_time true

before you start the editor in the terminal. Also comment out the start of the node in the bodenseeDataset.launch file.

About

ComputerVision for searchwing drones

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published