Skip to content

Latest commit

 

History

History
269 lines (227 loc) · 11.6 KB

README.md

File metadata and controls

269 lines (227 loc) · 11.6 KB
Raw

spinnaker_camera_driver: ROS driver for FLIR cameras based on the Spinnaker SDK

ROS driver for the FLIR cameras using the Spinnaker SDK.

NOTE: This driver is not written or supported by FLIR.

Tested cameras:

The following cameras have been used with this driver:

  • Blackfly S (USB3, GigE)
  • Blackfly (GigE)
  • Grashopper (USB3)
  • Chameleon (USB3, tested on firmware v1.13.3.00)

Note: if you get other cameras to work, please report back, ideally submit a pull request with the camera config file you have created.

Tested platforms

Software:

  • ROS2 Galactic under Ubuntu 20.04 LTS
  • ROS2 Humble under Ubuntu 22.04 LTS
  • Spinnaker 3.1.0.79 (other versions may work as well but this is what the continuous integration builds are using)

How to install

This driver can be used with or without installing the Spinnaker SDK, but installing the Spinnaker SDK is recommended because during its installation the USB kernel configuration is modified as needed and suitable access permissions are granted (udev rules). If you choose to not use the Spinnaker SDK, you must either run the linux setup script by running ros2 run spinnaker_camera_driver linux_setup_flir or perform the required setup steps manually. Without these setup steps, the ROS driver will not detect the camera. So you must either install the Spinnaker SDK (which also gives you the useful spinview tool), or follow the manual setup steps mentioned earlier.

Installing from packages

For some architectures and ros distributions you can simply install an apt package:

sudo apt install ros-${ROS_DISTRO}-spinnaker-camera-driver

The package will bring its own set of Spinnaker SDK libraries, so you don't necessarily have to install the SDK, but it's recommended, see above

Building from source

  1. Install the FLIR spinnaker driver. If you skip this part, the driver will attempt to download the Spinnaker SDK automatically to obtain the header files and libraries.
  2. Prepare the ROS2 driver build: Make sure you have your ROS2 environment sourced:
source /opt/ros/<my_ros_distro>/setup.bash

Create a workspace (~/ws), clone this repo:

mkdir -p ~/ws/src
cd ~/ws/src
git clone --branch humble-devel https://github.com/ros-drivers/flir_camera_driver
cd ..

To automatically install all packages that the flir_camera_driver packages depends upon, run this at the top of your workspace:

rosdep install --from-paths src --ignore-src
  1. Build the driver and source the workspace:
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_EXPORT_COMPILE_COMMANDS=ON
. install/setup.bash

Example usage

Single node launch

The driver comes with an example launch file (driver_node.launch.py) that you can customize as needed.

# launch with --show-args to print out all available launch arguments
ros2 launch spinnaker_camera_driver driver_node.launch.py camera_type:=blackfly_s serial:="'20435008'"

Stereo camera with synchronization

The launch file stereo_synced.launch.py provides a working example for launching drivers for two hardware synchronized Blackfly S cameras. It requires two more packages to be installed, cam_sync_ros2(for time stamp syncing) and exposure_control_ros2 (for external exposure control). The launch file also demonstrates how to use the driver as a composable node.

Features

The ROS driver itself has no notion of the camera type (Blackfly, Grasshopper etc), nor does it explicitly support any of the many features that the FLIR cameras have. Rather, all camera features (called Spinnaker Nodes) are mapped to ROS parameters via a yaml file that is specific to the camera. On startup the driver reads this parameter definition file. In the config directory there are some parameter definition files for popular cameras (blackfly_s.yaml etc) that expose some of the more frequently used features like frame rate, gain, etc. You can add more features by providing your own parameter definition file. The ROS driver code is just a thin wrapper around the Spinnaker SDK, and should allow you to access all features available in FLIR's spinview program. In addition to the parameters defined in the .yaml files, the driver has the following ROS parameters:

  • adjust_timestamp: see below for more documentation
  • acquisition_timeout: timeout for expecting frames (in seconds). If no frame is received for this time, the driver restarts. Default is 3s.
  • buffer_queue_size: max number of images to queue internally before shoving them into the ROS output queue. Decouples the Spinnaker SDK thread from the ROS publishing thread. Default: 4.
  • camerainfo_url: where to find the camera calibration yaml file.
  • compute_brightness: if true, compute image brightness and publish it in meta data message. This is useful for external exposure control but incurs extra CPU load. Default: false.
  • connect_while_subscribed: if true, connect to the SDK and pull data from the camera only while subscribers to image or meta topics are present. This feature reduces compute load and link utilization while no ROS subscribers are present, but adds latency on subscription in that the first image will be published up to 1s later than without this option.
  • dump_node_map: set this to true to get a dump of the node map. This
  • frame_id: the ROS frame id to put in the header of the published image messages.
  • image_queue_size: ROS output queue size (number of frames). Default: 4 feature is helpful when developing a new config file. Default: false.
  • parameter_file: location of the .yaml file defining the camera (blackfly_s.yaml etc)
  • serial_number: must have the serial number of the camera. If you don't know it, put in anything you like and the driver will croak with an error message, telling you what cameras serial numbers are available

Setting up GigE cameras

The Spinnaker SDK abstracts away the transport layer so a GigE camera should work the same way as USB3: you point it to the serial number and you're set.

There are a few GigE-specific settings in the Transport Layer Control group that are important, in particular enabling jumbo frames from the camera per FLIR's recommendations. The following line in your camera-specific config file will create a ROS2 parameter gev_scps_packet_size:

gev_scps_packet_size int "TransportLayerControl/GigEVision/GevSCPSPacketSize"

that you can then set in your ROS2 launch file:

 "gev_scps_packet_size": 9000

As far as setting up the camera's IP address: you can set up DHCP on your network or configure a static persistent IP using spinview in "Transport Layer Control">"GigE Vision". Check the box for "Current IP Configuration Persistent IP" first to enable it, then set your desired addresses under "Persistent IP Address", "Persistent Subnet Mask" and "Persistent Gateway". NOTE: these look like regular IPs, but to set them you have to enter the 32-bit integer representation of the IP address/mask. By hand/calculator: convert the IP octets from decimal to hex, then combine them and convert to a 32-bit integer, ex: 192.168.0.1 -> 0xC0A80001 -> 3232235521.

The "Transport Layer Control">"GigE Vision" section of spinview is also where you'll find that "SCPS Packet Size" setting, which you can change when not capturing frames, and verify it works in spinview and without needing to spin up a custom launch file to get started, though it helps, and you'll probably want one anyway to specify your camera's serial number.

For more tips on GigE setup look at FLIR's support pages here and here.

Time stamps

By default the driver will set the ROS header time stamp to be the time when the image was delivered by the SDK. Such time stamps are not very precise and may lag depending on host CPU load. However the driver has a feature to use the much more accurate sensor-provided camera time stamps. These are then converted to ROS time stamps by estimating the offset between ROS and sensor time stamps via a simple moving average. For the adjustment to work the camera must be configured to send time stamps, and the adjust_timestamp flag must be set to true, and the relevant field in the "chunk" must be populated by the camera. For the Blackfly S the parameters look like this:

    'adjust_timestamp': True,
    'chunk_mode_active': True,
    'chunk_selector_timestamp': 'Timestamp',
    'chunk_enable_timestamp': True,

When running hardware synchronized cameras in a stereo configuration two drivers will need to be run, one for each camera. This will mean however that their published ROS header time stamps are not identical which in turn may prevent down-stream ROS nodes from recognizing the images as being hardware synchronized. You can use the cam_sync_ros2 node to force the time stamps to be aligned. In this scenario it is mandatory to configure the driver to adjust the ROS time stamps as described above.

Automatic exposure

In most situations it is recommended to enable the built-in auto exposure of the camera. However, in a synchronized setting it is sometimes desirable to disable the built-in auto-exposure and provide it externally. For instance in a stereo setup, matching left and right image patches can be difficult when each camera runs its own auto exposure independently. The exposure_control_ros2 package can provide external automatic exposure control. To this end the driver publishes meta data messages and subscribes to camera control messages. See the launch file directory for examples.

How to add new features and develop your own camera configuration file

Check out this section for more information on how to add features.

Known issues

  1. If you run multiple drivers in separate nodes that all access USB based devices, starting a new driver will stop the image acquisition of currently running drivers. There is an ugly workaround for this currently implemented: if image delivery stops for more than acquisition_timeout seconds, the acquisition is restarted. This operation may not be thread safe so the driver already running could possibly crash. This issue can be avoided by running all drivers in the same address space with a composable node (see stereo launch file for example).

How to contribute

Please provide feedback if you cannot get your camera working or if the code does not compile for you. Feedback is crucial for the software development process. However, before opening issues on github first verify that the problem is not present when using spinview.

Bug fixes and config files for new cameras are greatly appreciated. Before submitting a pull request, run this to see if your commit passes some basic lint tests:

colcon test --packages-select spinnaker_camera_driver && colcon test-result --verbose

License

This software is issued under the Apache License Version 2.0. The file TargetArch.cmake is released under a custom license (see file)