Skip to content

Estimating Distance

Jump to bottom
Helen Burns edited this page Apr 15, 2024 · 5 revisions

Distance to from camera to cloud estimate.

Data

  • GPS of camera taken from metadata.

  • Yaw data recorded daily on camera set up (dubious quality data)

  • GOES optical depth data (regridded and parallax effect taken into account

Preprocessing

GOES data:

The optical depth data was adjusted and regridded following the method of Bielinski et al 2020, this is outline in script distance estimator parallax. Code provided my Declan Finney. For this workflow we extract area and times of interest from channel ABI-L2-CODC in the preprocessed data.

We extract the frames the cover the timeframe we have photos for . We then filter out thin cloud by setting an optical depth-threshold of 3.6, following the thin cloud threshold from ISCCP definitions. This primarily leaves cumulus congestus and deeper clouds.

Camera Field of View and location of objects

I order to work out what is in the picture we estimated the cameras field of view. Using thin lens equations [1,2]

$$\text{Horizontal FOV} = 2 \times \tan ^{-1} \left(\frac{ \text{sensor width} }{2 \times \text{focal length}} \right) $$

Estimating distance

Now we have the adjusted optical depth data and FOV information, the distance is estimated by using the camera location and YAW (plus or minus 10 deg to define a circle segment area in lat lon space. The max distance (i.e. radius) used is taken to be distance that covers the mountain.

We are assuming that the highest cloud in our images has the deepest optical depth in the satellite product. This works better in some cases than others, but in general the distance is fairly similar to most cloud in images for which sensible height estimates can be derived. We can now find the max optical depth within the camera FOV and calculate the distance using the Haversine formula to calculate the distance between two point on a sphere.

Below is an example plot:

https://github.com/cemac/DCMEX/blob/main/figures/FOV_example.png?raw=true

References:

  • Bielinski et al 2020
  • [1]"Fundamentals of Photonics, Bahaa E. A. Saleh and Malvin Carl Teich, 2007, FOV diagram"
  • [2]"Introduction to Modern Optics, Author: Grant R. Fowles, Dover Publications, 1989 - thin lens equations"

Gotchas/ things to think about

  • This was first attempted without cloud filtering, the filtering would have to be adjusted if looking at lower/higher clouds
  • parralax adjustment makes some height assumptions! So distance will still be aprroximate just more correct
  • Make sure lat lons used are same projections WGS84
  • Make sure to convert Degrees,minutes,seconds to decimal degrees our camera is in deg, mins sec.
  • Visualising the image indicated our YAW wasn't very accurate. Error estimated from matching up known objects, mountain ragne etc.
  • Resolution impacts the max height location
Clone this wiki locally