search

Extend-Robotics / camera_aravis

A ROS1 driver for GenICam based GigE and USB3 cameras.
Other
1 stars 1 forks source link

readme

camera_aravis

Actively maintained repository for the ROS1 camara_aravis driver. It is open source under the LGPL (like Aravis itself).

The Aravis library is a glib/gobject based library for video acquisition using Genicam cameras. It currently implements the gigabit ethernet and USB3 protocols used by industrial cameras.

The camera_aravis driver has long history of multiple forks and now abandoned GitHub repositories. This repository is based on https://github.com/florisvb/camera_aravis.git, which in turn was forked from a deleted github repo (https://github.com/CaeruleusAqua/camera_aravis), which was itself forked from https://github.com/ssafarik/camera_aravis.

Tested with Aravis version 0.8.27, requires >= 0.8.25 for multipart data handling.

The basic command to run camera_aravis:

$ rosrun camera_aravis cam_aravis

To run it in a given namespace:

$ ROS_NAMESPACE=cam1 rosrun camera_aravis cam_aravis

Continuous Integration

Service Noetic Master
GitHub build build
ROS Build Farm build N/A

Configuration

This ROS node publishes messages image_raw and camera_info for a specified camera. It supports a variety of camera features via the ROS reconfigure_gui, including the following:

Note that the above are also the ROS parameter names of their respective feature. You may set initial values for the camera by setting ROS parameters in the camera's namespace.

In addition to the above features, this driver now supports (almost) every feature of every camera, you just have to know how the feature is specified; each GenICam-based camera contains an XML file onboard, and by viewing this file you can determine which ROS parameters to set for camera_aravis to write to the camera. You can use arv-tool-0.2 to see the feature list and the XML file (e.g. "arv-tool-0.2 --name=Basler-21285878 features")

Note that for this special feature access, the ROS parameter type must match the feature type. For example, a Basler ac640 has a boolean feature called "GammaEnable", an integer feature called "BlackLevelRaw", and a string enum feature called "PixelFormat" that takes values (Mono8, Mono12, Mono12Packed, YUV422Packed, etc). The ROS params that you set for these must be, respectively, a bool, an integer and a string. Also note that boolean features must be specified as ROS params false/true, not as integer 0/1.

$ rosparam set /camera_aravis/GammaEnable false
$ rosparam set /camera_aravis/BlackLevelRaw 5
$ rosparam set /camera_aravis/PixelFormat Mono12
$ rosrun camera_aravis cam_aravis

camera_aravis supports multiple cameras, each of which may be specified on the command-line, or via parameter. Runs one camera per node.

To specify which camera to open, via the command-line:

$ rosrun camera_aravis cam_aravis _guid:=Basler-21237813

To specify which camera to open, via a parameter:

$ rosparam set /camera_aravis/guid Basler-21237813
$ rosrun camera_aravis cam_aravis

camera_aravis supports multisource cameras and multipart data

use channel_name, pixel_format, camera_info_url, frame_id to specify multisource/multipart camera

Parameter pixel_format_internal overrides pixel format used internally in camera_aravis.

You may use it to implement device quirks like custom device pixel formats that are sent under different GenICam/GiGE-Vision pixel format.

The pixel_format_internal works also with multisource and multipart devices

It supports the dynamic_reconfigure protocol, and once the node is running, you may adjust its parameters by running the following and then manipulating the GUI:

$ rosrun dynamic_reconfigure reconfigure_gui

There is an additional nice feature related to timestamps that unifies ROS time with camera time. We want a stable timestamp on the images that the camera delivers, giving a nice smooth time delta from frame to frame. If we were to use the ROS clock on the PC, by the time we get the image packets from the camera a variable amount of time has passed on the PC's clock due to variable network and system delays. The camera's onboard clock is stable but it doesn't match with the ROS clock on the PC, and furthermore since it comes from a different piece of hardware, the two clock's rates are slightly different.

The solution is to start with a base of ROS time, and to accumulate the dt's from the camera clock. To accomodate the difference in clock rates, a PID controller gently pulls the result toward ROS time.

Troubleshooting

MTU

For GigE Vision cameras in high bandwidth scenarios you may need to increase MTU for more fps.

Typically you would set matching MTU on host and camera but

Host MTU

On OS side you may do it with ip

# identify your network interface name
ip link list
# set jumbo frames 9000 mtu for network interface 'enp60s0'
sudo ip link set dev enp60s0 mtu 9000

Making the change permanent may vary depending on your OS and config.

If in doubt use GUI or TUI.

For Ubuntu 20.04 you may set it through NetworkManager:

Camera MTU

On device side you may do it with GenICam:

For example see photoneo_motioncam.launch

ARV_BUFFER_STATUS_TIMEOUT

If you see a lot of errors like

[ WARN] [1691651985.401198104]: (range_optical_frame (and possibly subframes)) Frame error: ARV_BUFFER_STATUS_TIMEOUT

Check if you are not getting errors on network layer

# check if errors are growing
netstat -us | grep errors

If yes try increasing network buffer sizes

# read current
sudo sysctl net.core.rmem_max
sudo sysctl net.core.rmem_default

The unit is bytes (kernel docs)

# set new, e.g. try 10x more than your original
sudo sysctl -w net.core.rmem_max=NEW_VALUE
sudo sysctl -w net.core.rmem_default=NEW_VALUE

Recheck if still getting errors.

Make final change permanent in /etc/sysctl.conf

tail /etc/sysctl.conf
# for what other values do
#kernel.sysrq=438

###################################################################

# Increase network buffer sizes
# 10 times my original max ~= 16 Mb

net.core.rmem_max = NEW_VALUE
net.core.rmem_default = NEW_VALUE