Python script example: Create a GPS tracking video from Overland data

[gloriouslyawkwardlife]

End result of this particular script: https://stevenbuehler.us/Data/OverlandVideo/2025-05-24.mp4

I did this for a specific day that I drove around the state (home to Kennedy Space Center to see a launch, to Daytona Beach, back home), but one can always use a ‘yesterday’/’today’ variable to replace the dates in the query string. My original purpose for doing this was to check the timestamps and make corrections to the data when I forget to choose to start or end a trip, and it kind of grew from there. I use a cgi Python script to parse the Overland GeoJSON payload into MySQL for efficiency, using UNIX epoch integers for timestamps in the data so as not to have to deal with timezone issues in the database itself.

The red dot indicating location is intended to be the same size as the horizontal_accuracy value from the Overland data, so I can see any error margins with the GPS fix. Most of the time the dot is 1 meter radius (best accuracy). There are a few spots in the video where the horizontal accuracy goes nuts (up to 5 or 6 km) and the whole frame turns red, and some ‘fluke’ frames from bad GPS readings. The video is 60 frames per second, one frame per GPS fix (so one minute of video = 1 hour of real time).

I wanted to do this with Wolfram Script (a little more robust), but I kept running into cloud resource limits; so Python it was.

https://stevenbuehler.us/now

---

#!/usr/bin/env python3.13

# TODO -
# - Add marker color for walk, car, bus, etc.

import requests, staticmaps, cv2, pandas, io, os, glob
from datetime import datetime, time, timedelta
from PIL import Image, ImageDraw

# Because Dreamhost doesn't let me access the database server outside of SSH I set up a
# Python cgi script to run the query and return headered CSV output, which I can then read into a Pandas
# dataframe.
data = requests.post(
    url="https://xxx.xxx/query.py",
    data={
        "query": f"SELECT timestamp, latitude, longitude, horizontal_accuracy from overland where timestamp >= {datetime.fromisoformat('2025-05-24T00:00:00-04:00').timestamp()} and timestamp < {datetime.fromisoformat('2025-05-25T00:00:00-04:00').timestamp()} and wifi != 'Buehler?' order by timestamp;"
    }
).content

# Pandas' read_csv requires a file or file-like object; thus the use of io.StringIO to make the incoming data appear like a file.
df = pandas.read_csv(io.StringIO(data.decode('utf-8')))

# The map drawing routine will choke without the following two lines
df['horizontal_accuracy']=df['horizontal_accuracy'].fillna(1)  # change any NaN values to 1 meter (default)
df['horizontal_accuracy']=pandas.to_numeric(df['horizontal_accuracy'],errors='coerce').astype(pandas.Int64Dtype()) # ensure all values are integer

# This section is needed as a workaround for the 'staticmaps' module, as recent versions of PIL deprecated the 'textsize' function
# that staticmaps relies on, so we create our own version.
def textsize(self: ImageDraw.ImageDraw, *args, **kwargs):
    x, y, w, h = self.textbbox((0, 0), *args, **kwargs)
    return w, h
ImageDraw.ImageDraw.textsize = textsize

# The actual map drawing routine. You can also use df.iterrows() in place of df.itertuples() and change all the row.x variables to row['x'] if you're a
# glutton for punishment; I just prefer dot-notation like JavaScript.

# Create a PNG image for each frame (database record).
for row in df.itertuples():
    map = staticmaps.Context()
    map.set_center(staticmaps.create_latlng(row.latitude,row.longitude))
    map.set_zoom(13)
    map.add_object(staticmaps.Circle(center=staticmaps.create_latlng(row.latitude,row.longitude),radius_km=row.horizontal_accuracy/1000))
    img = map.render_pillow(600,600)
    idraw = ImageDraw.Draw(img)
    idraw.text(xy=(20,20),text=f"{str(row.timestamp)}\n{datetime.fromtimestamp(row.timestamp).isoformat()}\n{(row.speed * 2.236936):.0f} mph",stroke_width=2,stroke_fill="black")
    img.save(f'/Users/steven/Documents/Slides/{row.timestamp}.png')

# use CV2 to compile the frames into a 60 fps video.
images = [img for img in sorted(os.listdir('/Users/steven/Documents/Slides')) if img.endswith('.png')]
frame = cv2.imread(os.path.join('/Users/steven/Documents/Slides',images[0]))
height, width, layers = frame.shape
video = cv2.VideoWriter('/Users/steven/Movies/2025-05-24.mp4', cv2.VideoWriter_fourcc(*'avc1'), 60, (width, height))
for image in images:
    video.write(cv2.imread(os.path.join('/Users/steven/Documents/Slides',image)))
cv2.destroyAllWindows()
video.release()

# Delete the PNG frames from the file system as they're not needed anymore.
for filename in glob.glob('/Users/steven/Documents/Slides/*.png'):
    os.remove(filename)

[lostfocus]

That's pretty neat! Thanks for sharing