ObjectDetector
Object Detection using OpenCV and ROS Interface
Table of Contents
Description
This package includes an object detection framework which can be used with ROS for any robot application you prefer.
Dependencies
- ROS Kinetic
- OpenCV 2.4.9
- ROS usb_cam driver (should also work with libuvc)
Status
The package has been tested with Ubuntu 16.04, ROS Kinetic and OpenCV 2.4.9 using the usb_cam driver package.
Install
- Install ROS: see instructions here
- Setup the ROS environment
- download and install OpenCV and cam driver packages.
- run
catkin build - source
source devel/setup.bash - start the usb cam driver, e.g.
rosrun usb_cam usb_cam_node - start the object detector using
rosrun color_detector color_detector - to see the output via ROS you could use a GUI e.g.
rqt_image_viewor direct line commandrosrun image_view image_view image:=/object_detector/image_treshold
Topics
Basically the object detector can either output the thresholded image or the contour image (or both if you want). Therefore, two topics are essential
/object_detector/image_treshold
outputs the tresholded image with parameters width=640, height=480, step=640, encoding="mono8"
/object_detector/contour_treshold
outputs the contour image with parameters width=640, height=480, step=640, encoding="mono8"
Troubleshooting
Always ensure that the usb_cam driver is running before you start the object detector node.