RealSense — UAV Vision Analytics#
A RealSense (or any UVC webcam) is a live camera in configs/cameras-pymavlink.yaml: the camera has a
device instead of a video file. The full settings are described in Configure the cameras;
this page is the short walk-through.
Prerequisites#
See System Requirements for the full list of software and hardware prerequisites.
Steps#
Find the color node. List the cameras and check the formats of the nodes:
lsusb v4l2-ctl --list-devices v4l2-ctl -d /dev/video4 --list-formats-ext # the color node offers YUYV or UYVY; depth is Z16, infrared GREY ls -l /dev/v4l/by-id/ # stable names; prefer them over /dev/videoN
Make it visible to the container. Put the path in
.env(make initfillsCAMERA_DEVICE_1with the first USB camera; up toCAMERA_DEVICE_4) and recreate the container:CAMERA_DEVICE_1=/dev/v4l/by-id/usb-Intel_R__RealSense_TM__Depth_Camera_435_..._<serial>-video-index0 make pymav-up
Enable the camera in
configs/cameras-pymavlink.yaml(it ships asrealsense_camera, disabled), with the same path asdevice:, then check and start:make check-cameras make start-rtsp # the cameras start when the vehicle is armed
Without a vehicle:
docker exec -it uav-va-dlstreamer-pipeline-server python3 /home/pipeline-server/scripts/pipeline_manager.py --start-now.View the stream at
rtsp://<HOST_IP>:8555/uav-realsense-camera(ffplay, VLC, or the ground-station dashboard, where the card has aLIVEbadge).
Pipelines#
Live cameras use uav_camera_<det> and uav_camera_<det>_<vlm>. configs/config-pymavlink.json has uav_camera_npu and
uav_camera_npu_gpu; for a CPU or GPU camera, copy the pipeline you need from
configs/config-pymavlink-realsense.json (all nine) into it and recreate the
container. make check-cameras says which pipeline is missing.
UAV Mission Compute SDK stack#
The SDK stack streams its cameras over RTSP and uses configs/cameras-uavsdk.yaml; see
Get started with the UAV Mission Compute SDK.