optical_code_utils ROS package

Read optical codes (QR codes, Barcodes, and many more) within ROS in 2d and 3d input streams in ROS1 and ROS2

ROS nodes to:

• read the values stored by a code seen in a video stream
• get the transform between a code and the origin of a point cloud or depth image video stream (TODO)
• perform the same utilities on single images

Getting optical code values from sensor_msgs/Image

Video streams in ROS are represented as a series of published images (sensor_msgs/Image). It would be computationally expensive to run the QR code recognizer on every single frame in the video stream. Instead, the node provides these alternative methods of interaction:

1. Always have the most recent frame stored, so that a service can query for the values within currently-visible QR codes
2. Constantly publish the currently-visible QR codes, but at a very limited rate (e.g., 1Hz)
3. Allow the user of the service to provide the image, and receive the resulting strings

One node handles all three cases:

rosrun qr_utils qr_reader_2d

The node will attempt to subscribe to a stream of sensor_msgs/Image on /input_image, you should be able to call the service /qr_reader/read_environment. The service takes std_msgs/Empty as its input, so there's nothing to input. This returns std_msgs/string[], where each entry in the array is an individual QR code in the input image.

To make the node constantly publish QR code observations, the node will again need to be subscribed to an input image stream. With the parameter /qr_reader/publish_hz set to non-zero, however, readings will be published at the rate specified. With a negative Hz or a Hz exceeding that of the input, the publisher Hz will be set to the input Hz. Note, however, that this is very rarely necessary, and given the increased computational cost, it is not recommended. As with the service previously described, the output will be an array of strings.

Finally, an image can be provided which the reader will attempt to parse. This is done via the /qr_reader/read_image service, which takes in a sensor_msgs/Image and returns an array of strings. If this is your use-case and you would find it useful to turn off the Image subscriber that isn't being used in this use-case, it can be disabled via the qr_reader/listen_to_image boolean.

Getting optical code values from depth images

rosrun qr_utils qr_reader_3d

Similar to previously, this will return all visible codes in an environment. However, it will also publish the transform between the viewed codes and the optical frame.

See Also

The zbar_ros package predates this package and provides similar functionality by wrapping the zbar library as a ROS package. This package differs from that on in that this package supports ROS2, provides functionality for dealing with point clouds, and has a slightly different interface. However, this repo learns a lot from the zbar_ros package, and the zbar_ros package is within the ROS Device Drivers group, so it's probably more stable and better supported.

Contributions Welcome