Jetson and Rekor setup
Configure the local ingest service, Scout agent, camera stream and display settings.
Scope and preparation
This guide follows the Rekor installation and configuration sections of LPR - Jetson - Rekor Setup. It covers a Jetson running the Rekor Scout agent and a local ingest relay. Confirm the Jetson, camera and target lane middleware addresses from the site record before configuring them. Use Chrome or Chromium for browser access.
The documented data path is camera stream → Rekor Scout → local ingest relay → lane middleware. The local relay endpoint is localhost:5000/rekor-ingest.
Set up the ingest relay
- Open a terminal on the Jetson with Ctrl + Alt + T.
- Place the approved
rekor-ingest.pyfrom the NXT Wash Middleware-Installer repository in the intended service user's home directory. - Set the script's
SET_LATEST_CAR_APIto the assigned lane middleware endpoint supplied for this installation. Confirm the destination IP belongs to that lane. - Install the documented dependency with
pip3 install flask. - Create
/etc/systemd/system/rekor-ingest.serviceusing the service structure below. Replace both home-directory paths and the User value with the actual service account; the placeholder is not a literal username.
[Unit]
Description=Rekor Ingest Relay
After=network.target
[Service]
ExecStart=/usr/bin/python3 /home/<jetson-user>/rekor-ingest.py
WorkingDirectory=/home/<jetson-user>
Restart=always
RestartSec=5
User=<jetson-user>
Environment=PYTHONUNBUFFERED=1
[Install]
WantedBy=multi-user.target
The following installation commands register the service, enable it at boot and restart it. Run them after saving the completed file:
sudo systemctl daemon-reload
sudo systemctl enable rekor-ingest.service
sudo systemctl restart rekor-ingest.service
sudo systemctl status rekor-ingest.service
Check that the service starts successfully before configuring Scout to send events to it.
Install and connect Rekor Scout
Use the OpenALPR installer with its install_agent, install_nvidia and register_agent selections. Use the approved installer for the deployed Jetson image and the site's assigned Rekor license. Complete the agent, NVIDIA and OpenALPR configuration stages described by that installer.
In the agent configuration wizard, choose Other HTTP Web Server, enter localhost:5000/rekor-ingest, then select Apply and Next. Resolve any reported error before continuing.
Close the initial Add Camera dialog and Scout. Apply the approved NXT Wash configuration to /etc/openalpr/alprd.conf, then reboot as directed by the setup note. On the desktop, right-click rekor-scout-agent.desktop, choose Allow Launching, and open Scout.
Agent settings
Compare these GPU acceleration and startup settings with the configuration for the installed agent:
| Setting | Documented value |
|---|---|
hardware_acceleration |
1 |
gpu_batch_size |
5 |
auto_start_on_boot |
1 |
upload_data |
1 |
upload_address |
localhost:5000/rekor-ingest |
country |
us |
classify_vehicles |
1 |
motion_detection |
0 |
store_video |
0 |
In Configure → Display Settings, set Display Buffer to 0 seconds and Show Motion to False. Set Show Video, Highlight Plates, Show Plate Number, Show Province, Show Detection Mask, and Show Vehicles to True.
Add the lane camera
Choose Add → IP Camera (Manual). Use the correct camera IP and its assigned access details. Use the matching stream path:
| Camera example | Stream address template |
|---|---|
| Axis | rtsp://<camera-IP>:554/axis-media/media.amp?videocodec=h264&streamprofile=nxtwash |
| Micrologic | rtsp://<camera-IP>/media/video1 |
| ICS | rtsp://<camera-IP>:554/cam0_0 |
For Axis, confirm that the nxtwash profile exists and uses H.264; see Axis camera configuration. Replace the example address with the assigned camera IP.
With the camera online, select Test. After a successful image test, save the camera with a recognizable site and lane name and select Finish. Open Configure → Detection Mask and apply the approved lane inclusion area.
Verify the assigned lane
Confirm the saved camera opens in Scout, the ingest service remains running, and the event destination is the intended lane middleware. Use LPR architecture and lane mapping for the camera-to-lane checks. A successful stream test establishes camera access; the lane result also needs verification.