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_view or direct line command rosrun 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.