magnetic_block_detection

This system was developed to detect and locate wooden magnetic blocks within a region bounded by at least three Alvar AR tags.

Composed of two packages:

• block_detector: Block detection algorithm
• ar_track_alvar: Forked (and modified) Alvar AR package

System

The Alvar subsystem is a modified verison of the official package. The only change made is that this version also outputs the 2D image coordinates of the markers on a separate topic. This, in turn, allows for our code to produce a 2D to 3D mapping with the only assumption being the markers are placed on a flat surface. This version of the package should be drop-in compatible with other uses. Please refer to the package's README for more details.

Our package (block_detector) provides the 2D to 3D mapping code, simple computer vision block detector, and calibration code. We also have an experimental multi-camera pipeline (though it is still in development). Please refer to the package's README for more details.

Note, block_detector already has the following python modules within the package:

• christophhagen/averaging-quaternions ~ Please see the README.

Install

Make sure to install the necessary camera ROS drivers. For instance if using a USB webcam then install usb_cam.

Then install the following python modules:

• scikit-learn
  • numpy
  • matplotlib
• cv2
• shapely
• yaml

Run

Enter following into terminal:

roslaunch block_detector main.launch video_src:=<video-device e.g., /dev/video0>