Camera Telemetry

Overview

Streams live camera video to Serial Studio over UDP as fast as possible. Only FPS and frame count are sent alongside the image.

No microcontroller needed, a standard webcam or built-in laptop camera is all that's required. A synthetic test pattern is available when no camera is connected (--no-camera).

What It Streams

Image data (Image View widget)

Raw JPEG frames sent directly over UDP. The Image View widget uses autodetect mode: it finds frames by their FF D8 FF magic bytes and FF D9 end-of-image marker, independently of the CSV telemetry parser.

Telemetry (Performance group)

Frame Format

Two types of data share the same UDP byte stream:

JPEG image frame — raw bytes, no additional framing. Autodetected by magic bytes.

CSV telemetry frame — wrapped in a 3-byte start sentinel and 2-byte end sentinel:

AB CD EF  fps,frame_count  FE ED

Example (hex):

AB CD EF 32 39 2E 38 2C 31 30 32 34 FE ED
         2  9  .  8  ,  1  0  2  4

The sentinels are chosen so they are statistically impossible to occur in JPEG, PNG, BMP, or WebP compressed data. AB CD EF is not a valid JPEG marker sequence and does not arise from JPEG byte-stuffing rules. The FrameReader extracts only bytes between the two sentinels; the ImageFrameReader ignores ASCII telemetry packets since they carry no image magic bytes.

How to Run

Step 1 — Install dependencies

pip install opencv-python

Step 2 — Start the script

python3 camera_telemetry.py

Options:

python3 camera_telemetry.py --no-camera          # no hardware needed
python3 camera_telemetry.py --camera 1           # secondary camera
python3 camera_telemetry.py --fps 60             # push higher rate
python3 camera_telemetry.py --quality 60         # smaller packets

Step 3 — Configure Serial Studio

1. Open Serial Studio and load Camera Telemetry.ssproj.
2. In the Setup panel set Bus Type → Network Socket, Socket Type → UDP.
3. Set Port to 9000 (or whatever --port you used).
4. Click Connect.

Performance Tips

• Lower --quality reduces JPEG size and UDP fragmentation pressure. Quality 60–75 is usually sufficient for 640×480 preview.
• Frames wider than 640 px are automatically downscaled before encoding.
• The synthetic pattern (--no-camera) animates a colour gradient with a moving circle — useful for testing without any camera hardware.

Dependencies

• Python 3.8 or later
• opencv-python — pip install opencv-python

License

Copyright (C) 2020-2025 Alex Spataru SPDX-License-Identifier: GPL-3.0-only OR LicenseRef-SerialStudio-Commercial