dji_parse

A simple utility for parsing telemetry data from MP4 recordings generated by DJI drones - such as the DJI Mini 2. Output data can be stored as either CSV, JSON, or GPX data.

Requirements

1. Python 3.11 or above.
2. Poetry for Python dependency management.
3. ffmpeg installed locally on your system.
4. Recordings from a DJI Drone with subtitle telemetry enabled.

Usage

➜  dji_parse git:(main) ✗ poetry run dji_parse --help
Usage: dji_parse [OPTIONS] INPUT_VIDEO OUTPUT

  Parse INPUT_VIDEO file for embedded DJI telemetry, writing data to OUTPUT.

  Supports writing telemetry data in either CSV, JSON, or GPX format. Defaults
  to GPX.

Options:
  --output-format [csv|gpx|json]
  --help                          Show this message and exit.

Output

Telemetry samples include:

1. A timestamp computed by adding the sample timestamp (see (2)) to the video's creation date. (string in ISO format)
2. A relative timestamp. (HH:MM:SS)
3. A Latitude and Longitude pair. (float)
4. Relative Elevation. (float in unit meters)
5. Distance. (float in unit meters)
6. Horizontal and Vertical Velocities. (float in unit meters/secs)

From the samples I've parsed it appears that the telemetry subtitles are generated at a sample-rate of 1Hz - i.e. they provide a single observation per second - with GPS precision of 4 decimal places.

Better values are available from the FlightRecord binary logs, however the format isn't documented and there are issues with the DJI supplied SDK/parser.

Example: CSV

time,timestamp,latitude,longitude,rel_elevation,distance,horizontal_velocity,vertical_velocity
...
2022-12-24T11:34:55+00:00,00:01:46,51.3847,0.1272,-0.5,1.91,0.0,-0.0

Example: GPX

<?xml version="1.0" encoding="UTF-8"?>
<gpx xmlns="http://www.topografix.com/GPX/1/1" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.topografix.com/GPX/1/1 http://www.topografix.com/GPX/1/1/gpx.xsd" version="1.1" creator="gpx.py -- https://github.com/tkrajina/gpxpy">
  <trk>
    <trkseg>
      <!-- ... -->  
      <trkpt lat="51.3847" lon="0.1272">
        <ele>-0.5</ele>
        <time>2022-12-24T11:34:55Z</time>
      </trkpt>
    </trkseg>
  </trk>
</gpx>

Example: JSON

[
  {
    "time": "2022-12-24T11:34:55+00:00",
    "timestamp": "00:01:46",
    "latitude": 51.3847,
    "longitude": 0.1272,
    "rel_elevation": -0.5,
    "distance": 1.91,
    "horizontal_velocity": 0.0,
    "vertical_velocity": -0.0
  }
]

Thanks

This wouldn't have been possible without ewpratten/dji3d - and if you're looking to 3D plots from the extracted data then it's probably the better option!

Before seeing @ewpratten's dji3d repository I was trying to parse the subtitles out manually - which meant a lot of MP4 file format headaches.

License

This project is made available under the terms of the The MIT License.