mjpye ▮

Reolink camera ROS2 interface

ROS2 interface

There is a package which defines actions for a PTZ camera, so a generic interface which I should use. ptz_action_server. It doesn’t include a driver for working with my E1 zoom camera though from what I see, so I should use this package: reolinkapipy

ONVIF - not used

Tried this ONVIF python package and it works, but gives nothing over the Python on in terms of absolute position. This is because the PTZ doesn’t seem to be capable. Remember you have to enable ONVIF in reolink app.

(venv) rpi@rpi-desktop:~/reolinkapi_testing$ onvif-cli -u 'admin' -a '<password>' --host '192.168.50.2' --port 8000 --wsdl python-onvif-zeep/wsdl/

The above starts the interactive session, from which we can run commands like below.

ONVIF >>> cmd ptz GetConfigurationOptions {'ConfigurationToken': '000'}

According to API though we can go to a preset value, so not sure how this works without absolute positioning.

Preset values

If we wanted to do missions in the future with the PTZ camera, it would be necessary to save the orientation the camera is looking. Sadly we don’t seem to be able to pull absolute data about camera position. This is strange though, as we can set “presets” with the API or App and go back to them. It will change Pan/Tilt/Zoom to get back to the original preset. At least if I want this feature, I can just map lots of camera poses and save them as presets, or whenever adding a POI save that as a preset to return to. Likely limited on max number of preset values though. According to docs is 64. API docs Need to test that presets save correctly when I move the camera with API and not the App.

Creating the ROS2 package

cd /home/rpi/ros2_ws/src

ros2 pkg create --build-type ament_python --node-name reolink_cam_ros2_node reolink_cam_ros2

cd /home/rpi/ros2_ws
colcon build --packages-select reolink_cam_ros2

Then to test it worked try running.

source /home/rpi/ros2_ws/install/setup.bash
ros2 run reolink_cam_ros2 reolink_cam_ros2_node

Which gives the Hello World output.

Continue the tutorial from here.

For tab completion:

ros2 topic pub /joy sensor_msgs/msg/Joy "h

then hit tab.

To use left bumper to trigger looking around:

sensor_msgs.msg.Joy(header=std_msgs.msg.Header(stamp=builtin_interfaces.msg.Time(sec=0, nanosec=0), frame_id=''), axes=[0.0, 0.0, 0.0, 0.0, 0.0, 0.0], buttons=[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])

When you have Python dependencies for the ROS2 package add them to the setup.py file like so:

install_requires=[
	'setuptools',
	'reolinkapi>=0.1.5',
	'pyyaml',
	],

rclnodejs adding times to messages

Sometimes it is necessary to have timestamps on the messages to check we are not acting on old information. Here is how to add it, and the full example is in thewebrtc-to-ros2 node:

clock = node.getClock();
const now = clock.now();
const {seconds, nanoseconds} = now.secondsAndNanoseconds

const joyMessage = {
	header: {
		stamp: {
			sec: seconds,
			nanosec: nanoseconds
		},

Have added Joy message processing logic to reolink_cam_ros2_node.py, now just need to actually call the reolink API inside functions like reolink_move_callback.

How does the node work

We connect to the camera using reolinkapi and a reolink_camera.yaml config file with IP, Username and Password.

So we listen to Joy commands on the Joy topic, and depending on the value of sticks and buttons (only when LB is pressed), we publish to some other topics.

/Joy >> /reolink_cam/ptz/move OR zoom OR focus

These topics have callbacks which call the reolinkapi with commands like:

self.reolink_ptz.move_left(20)

These commands run forever until preempted by:

self.reolink_ptz.stop_ptz()

We preempt whenever:

  • LB is not pressed
  • RB is pressed
  • No conditions for move, zoom or focus are met

We only send the stop command once, just like we only send each move/zoom/focus commands once. We do this by checking what the last sent message was.

reolinkapi and Python version

get_ptz_presets() is failing as it is not included in the current versions I have installed.

reolinkapi==0.1.5
python3==3.10.12

Therefore we need reolinkapi==0.4.1 but that requires:

python==3.12.4

Made a pull request to fix the broken add_preset function here

Testing from MacBook with latest version of reolinkapi. Edit files here to see changes:

/Users/matthewpye/Documents/reolink_api_testing/venv/lib/python3.12/site-packages/reolinkapi/mixins

Already tried moving to presets and setting presets from the MacBook and it works. Get the hang of that then test it from the robot instead.

Testing changed file on rpi

Maybe I can just replace the ptz.py file with the new one 0.4.1 and it works on the rpi. Made mp_archive here then replaced with 0.4.1 to test:

/home/rpi/.local/lib/python3.10/site-packages/reolinkapi/mixins

Testing and it actually seems like we can just force Pip to not care about the python version requirement by installing like this:

pip3 install --ignore-requires-python reolinkapi==0.4.1

So the only problem now is that reolinkapi==0.4.1 has the bug where when you call:

response = self.reolink_ptz.add_preset(2, "Left")

It fails because of the missing parameter that should be fixed in the Pull request. For now I will edit the file manually. So the file located here is fixed on my rpi:

/home/rpi/.local/lib/python3.10/site-packages/reolinkapi/mixins/ptz.py

Note: Just pay attention to the PTZ. It seemed to turn off once today and re-calibrate. Maybe was drawing too much current and the Step down converter max current needs to be adjusted.

May 25, 2025 ∙