Instead of starting Alphasense with the standard configuration, the driver can be given a custom settings file. An example is provided here: ncamera_settings.yaml. This can be passed to the driver with one of the commands shown below.
Standalone viewer:
viewalphasense --ncamera-settings /path/to/ncamera_settings.yamlROS driver:
rosrun alphasense_driver_ros alphasense_driver_ros _ncamera_settings:=/path/to/ncamera_settings.yamlWhen using the ROS driver, it is possible to change the settings dynamically as an alternative to providing a settings file. To adjust the configuration of the driver, use rqt_reconfigure after launching the driver.
rosrun rqt_reconfigure rqt_reconfigure
This application lists all availabe settings and allows to change them on the fly. Those are described in more detail in the sections below.
Each camera can have distinct settings for (auto) exposure and gain. The exposure intervals of all cameras will be temporally aligned in the center.
The exposure time of a camera can be fixed by setting the
auto_exposure_enabled parameter to false and setting the
manual_exp_time_us to the desired exposure time in microseconds.
Alphasense is equipped with an intelligent auto exposure algorithm which is specially designed for computer vision and state estimation applications. It detects areas in the image that contain high gradients and adjusts the exposure time accordingly. As a result, it is possible to recognize structure in the environment under the influence of sunlight, for example.
Auto exposure can be enabled by setting the auto_exposure_enabled parameter
to true. The algorithm adjusts the exposure time until it reaches the overall
image brightness defined in the autoexp_target_median setting. The target
brightness is encoded using 8-bit grayscale: 0 representing black and 255
white. The minimum and maximum exposure time range within which the algorithm
can operate can be configured with min_exp_time_us and max_exp_time_us
respectively. Both parameters are given in microseconds. If autogain is
enabled, the gain will be increased if the maximum exposure time is too low in
order to reach the desired image brightness. The parameter autoexp_speed
adjusts how smooth the change in exposure time and consequently image
brigthness is. However, when autoexp_speed is reduced, convergence is slowed
down.
To make matching across stereo cameras more stable it helps to equalize the exposure of both cameras. This can be done in the Alphasense by running auto exposure for only one camera in the pair, and copy the result to the other camera.
A camera can be configured to copy the exposure parameters of another camera with the autoexp_master_camera_index parameter.
ncamera:
# ...
cams:
0:
auto_exposure_enabled: true
# ...
1:
autoexp_master_camera_index: 0The auto exposure algorithm can be configured to only optimize brightness of a certain rectangular patch of the image.
The boundaries of the patch can be set as follows:
autoexp_start_row: 0
autoexp_end_row: -1
autoexp_start_col: 0
autoexp_end_col: -1For autoexp_end_row and autoexp_end_col, -1 equals the maximum height and
width. So in this case, the whole image is being used.
The camera gain parameter can be used to adjust light sensitivity of the
sensor. Increasing the gain will lead to brighter images with shorter exposure
times. Note that this comes at the cost of increased noise in the image.
Setting manual_gain to 0 will result in no amplification and 255 will
increase sensitivity by a factor of 5.6. If auto exposure is enabled, this
parameter has no influence.
This setting is only relevant if auto exposure is enabled. Enabling this
parameter allows the auto exposure algorithm to increase the camera gain when
maximum exposure time is reached. This enables operation in very dark
environments or when maximum exposure time is too low to reach a sufficiently
bright image. In case the desired image brightness can be achieved adjusting
exposure time only, gain is kept at the lowest level to reduce noise. The
autoexp_max_gain allows for manual control over the maximum gain value, the
algorithm can apply.
⚠️ Breaks factory calibration: Do not rotate the camera images when using the factory supplied calibration!
The camera image can be rotated by 180 degrees. This can be done by setting the
rotate_image parameter to true. The rotation of the images happens
directly on the Alphasense camera main board.
The number of enabled cameras can be set with the num_cams option. The camera
ports until the num_cams value will be enabled. By disabling camera ports, the
maximum allowed frame rate of the enabled cameras can be increased.
The camera frame rate can be set with the camera_frequency_hz option. The IMU
frequency can be set to 100, 200 or 400 Hz with the imu_frequency_hz option.
The minimum frame rate is 1Hz. The maximum frame rate depends on the number of cameras,
the maximum/manual exposure time, and network tuning parameters
(inter_packet_delay_us/pixels_per_packet). The table below gives the absolute
maximum frame rates for each number of cameras with exposure at 10us, inter_packet_delay_us at 1.0 us and pixels_per_packet at 7200.
| Number of Cameras | Core Research (0.4 MP, Mono) | Core Research (1.6 MP, Mono) |
|---|---|---|
| 1 | 75 | 30 |
| 2 | 75 | 30 |
| 3 | 75 | 24 |
| 4 | 72 | 18 |
| 5 | 58 | 14 |
| 6 | 48 | 12 |
| 7 | 41 | 10 |
| 8 | 35 | 8 |
The sensitivity of the gyroscope can be set with the gyro_range_deg_per_sec.
The parameter sets the maximum measurable rotation rate. Setting this value
higher will decrease the gyro sensitivity.
The peak bandwidth used by the Core Research can be set with the peak_bandwidth_limit_mbps parameter.
The Core Research packet size can be set with the pixels_per_packet option.
It is best to increase the pixels_per_packet to the highest possible allowed by your network setup (The maximum is 7200).
ℹ️ Info: The MTU of the network card to which the Core Research is connected needs to be at least
pixels_per_packet+ 60, otherwise the driver cannot receive the image stream packets. In that case you will get "Image receive timed out." errors.
See Maximize network performance for more information.
The driver can be configured to not stream IMU and camera frames before the device successfully
synchronizes its clock. This prevents a big timestamp jump in the streams.
To enable this set expect_ptp_time_synchronization to true, by default this is set to false.
To prevent timestamps from drifting apart when PTP synchronization is lost, a timeout can be set after which the driver
blocks the streams. This timeout in seconds is configured with the max_ptp_sync_lost_time_s parameter.
By default this parameter is set to 60 seconds. The timeout mechanism can be disabled by setting this to 0.